Papers with Explainability & Interpretability in NLP
NLP+Vis: NLP Meets Visualization (2023.emnlp-tutorial)
Copied to clipboard
| Challenge: | This tutorial will introduce NLP+Vis with a focus on two main threads of work: NLP for Vis and Vis for NLP. |
| Approach: | tutorial will introduce NLP+Vis with a focus on two main threads of work . overview of research topics on combining NLP and Vis techniques will be covered . |
| Outcome: | The tutorial will introduce NLP+Vis with a focus on two main threads of work . it will provide an overview of research topics on combining NLP and Vis techniques . |
Interpreto: An Explainability Library for Transformers (2026.acl-demo)
Copied to clipboard
Antonin Poché, Thomas Mullor, Gabriele Sarti, Frédéric Boisnard, Corentin Friedrich, Charlotte Claye, Francois Hoofd, Raphael Bernas, Nicholas Asher, Celine Hudelot, Fanny Jourdan
| Challenge: | Interpreto is an open-source Python library for interpreting HuggingFace language models . it provides attribution methods and concept-based explanations . documentation or metrics are sometimes missing due to the complexity of the pipeline . |
| Approach: | Interpreto is an open-source Python library for interpreting HuggingFace language models . it provides attribution methods and concept-based explanations . authors welcome issues and pull requests . |
| Outcome: | Interpreto is an open-source Python library for interpreting HuggingFace language models . it provides attribution methods and concept-based explanations . the library welcomes issues and pull requests . |
AllenNLP Interpret: A Framework for Explaining Predictions of NLP Models (D19-3)
Copied to clipboard
| Challenge: | Existing interpretation codebases make it difficult to apply these methods to new models and tasks. |
| Approach: | They propose a framework for interpreting NLP models that provides explanations for specific models. |
| Outcome: | The proposed framework provides interpretation primitives for any AllenNLP model and task, a suite of built-in interpretation methods, and a library of front-end visualization components. |
Explanation in the Era of Large Language Models (2024.naacl-tutorials)
Copied to clipboard
| Challenge: | Explanation has long been a part of communication, where humans use language to elucidate each other and transmit information about mechanisms of events. |
| Approach: | They review the opportunities and challenges of explanations in the era of large language models and examine how they can be used to generate explanations. |
| Outcome: | The proposed methods are based on the models of large language models (LLMs) and their opaque nature. |
Transformer-specific Interpretability (2024.eacl-tutorials)
Copied to clipboard
| Challenge: | Transformers are dominant play-ers in various scientific fields, but their inner workings remain opaque. |
| Approach: | This tutorial presents a trending approach to interpreting Transformers . it uses specific features of the Transformer architecture to quantify context- mixing interactions . |
| Outcome: | This tutorial aims to show how a new trending approach can be applied to Transformer-based models. |
Explainable CED: A Dataset for Explainable Critical Error Detection in Machine Translation (2024.naacl-srw)
Copied to clipboard
| Challenge: | Existing studies of critical error detection lack content addressing the causes of catastrophic errors. |
| Approach: | They propose a dataset that introduces the attributes of error explanation and correction regarding critical errors. |
| Outcome: | The proposed dataset reduces time costs and mitigates human annotation bias. |
Don’t Shoot The Breeze: Topic Continuity Model Using Nonlinear Naive Bayes With Attention (2024.emnlp-industry)
Copied to clipboard
| Challenge: | Large-scale language models (LLMs) are becoming increasingly popular in business scenarios, but maintaining topic continuity is a challenge. |
| Approach: | They propose a topic continuity model that assesses whether a response aligns with the initial conversation topic using a Naive Bayes approach. |
| Outcome: | The proposed model outperforms existing models in handling lengthy and complex conversations. |
Contrastive Explanations of Text Classifiers as a Service (2022.naacl-demo)
Copied to clipboard
| Challenge: | ContrXT provides time contrastive explanations of black box text classifiers by manipulating binary decision diagrams. |
| Approach: | They propose a system that provides time contrastive explanations of black box classifiers as a service by manipulating binary decision diagrams. |
| Outcome: | The proposed system has a throughput of 2.55 users per second and is available as a python pip package. |
Attention Flows are Shapley Value Explanations (2021.acl-short)
Copied to clipboard
| Challenge: | Shapley Values are a popular type of explanation in machine learning, but leave-one-out and attention-based explanations still predominate in NLP. |
| Approach: | They propose to use attention flow to explain the importance of features, embeddings, and even neurons to explain credit assignment problems in cooperative game theory. |
| Outcome: | The proposed explanations can explain the importance of features, embeddings, and even neurons, but in NLP, leave-one-out and attention-based explanations still predominate. |
Simplifying Outcomes of Language Model Component Analyses with ELIA (2026.eacl-demo)
Copied to clipboard
| Challenge: | ELIA is an interactive web application that simplifies the outputs of various language model component analyses for a broader audience. |
| Approach: | They propose to use a vision-language model to automatically generate natural language explanations for the complex visualizations produced by these methods. |
| Outcome: | The proposed system integrates three key techniques and generates natural language explanations for complex visualizations. |
WAX: A New Dataset for Word Association eXplanations (2022.aacl-main)
Copied to clipboard
| Challenge: | Word associations are among the most common paradigms to study the human mental lexicon. |
| Approach: | They present a large dataset of word associations with explanations and relation labels . they show that current language models struggle to capture the diversity of human associations . |
| Outcome: | The proposed model fails to capture the diversity of human associations, the authors show . they show that the model is a rich benchmark for commonsense modeling and generation. |
Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals (2021.tacl-1)
Copied to clipboard
| Challenge: | Amnesic probing is a method that focuses on how information is being used, rather than on what information is encoded. |
| Approach: | They propose a method that focuses on how the information is being used rather than on what information is encoded. |
| Outcome: | The proposed method is based on a BERT dataset to ask questions that were not possible before . it shows that probing performance is not correlated to task importance . |
A Hypothesis-Driven Framework for the Analysis of Self-Rationalising Models (2024.eacl-srw)
Copied to clipboard
| Challenge: | Recent advances in LLMs generating longer coherent text have popularised self-rationalising models, which produce a natural language explanation alongside their output. |
| Approach: | They propose a Bayesian network-based hypothesis-driven statistical framework that allows us to judge how similar LLM-generated free-text explanations are to LLMs. |
| Outcome: | The proposed framework does not exhibit a strong similarity to GPT-3.5. |
Red Dragon AI at TextGraphs 2019 Shared Task: Language Model Assisted Explanation Generation (D19-53)
Copied to clipboard
| Challenge: | The TextGraphs-13 Shared Task on Explanation Regeneration asked participants to develop methods to reconstruct gold explanations for elementary science questions. |
| Approach: | The TextGraphs-13 Shared Task on Explanation Regeneration asked participants to develop methods to reconstruct gold explanations for elementary science questions. |
| Outcome: | The Explanation Regeneration shared task asked participants to develop methods to reconstruct gold explanations for elementary science questions. |
Inference to the Best Explanation in Large Language Models (2024.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have found success in real-world applications, but their underlying explanatory process is still poorly understood. |
| Approach: | They propose to use a framework inspired by philosophical accounts on Inference to the Best Explanation (IBE) to advance the interpretation and evaluation of LLMs’ explanations. |
| Outcome: | The proposed framework can identify the best explanation with up to 77% accuracy (27% above random) while being intrinsically more efficient and interpretable. |
This Patient Looks Like That Patient: Prototypical Networks for Interpretable Diagnosis Prediction from Clinical Text (2022.aacl-main)
Copied to clipboard
Betty van Aken, Jens-Michalis Papaioannou, Marcel Naik, Georgios Eleftheriadis, Wolfgang Nejdl, Felix Gers, Alexander Loeser
| Challenge: | a novel method for diagnosis prediction from clinical text is needed in clinical practice . prototypical part networks and label-wise attention are used to make models interpretable and helpful . |
| Approach: | They propose a deep neural model that makes predictions based on parts of the text that are similar to prototypical patients. |
| Outcome: | The proposed method outperforms baseline models on two clinical datasets and provides valuable explanations for clinical decision support. |
Interpreting Language Models with Contrastive Explanations (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing explanation methods conflate evidence for various features to predict a token . existing explanation methods are less interpretable for human understanding . |
| Approach: | They propose to explain language models contrastively by looking for salient input tokens that explain why the model predicted one token instead of another. |
| Outcome: | The proposed explanations are better than non-contrastive explanations for language models . they show that contrastive explanations improve simulability for human observers . |
Leveraging Human Production-Interpretation Asymmetries to Test LLM Cognitive Plausibility (2025.acl-short)
Copied to clipboard
| Challenge: | Existing research on the linguistic capabilities of large language models has focused on their performance in language interpretation. |
| Approach: | They examine whether large language models (LLMs) process language similarly to humans . they use an empirically documented asymmetry between production and interpretation in humans a testbed . |
| Outcome: | The proposed model can replicate human-like distinctions between production and interpretation. |
Unification-based Reconstruction of Multi-hop Explanations for Science Questions (2021.eacl-main)
Copied to clipboard
| Challenge: | Existing approaches build explanations considering each question in isolation, but new approach leverages explanatory patterns emerging in scientific explanations. |
| Approach: | They propose a framework for reconstructing multi-hop explanations in science Question Answering . they integrate lexical relevance with the notion of unification power to rank atomic facts . |
| Outcome: | The proposed method achieves results competitive with Transformers, but is faster and scalable to large explanatory corpora. |
Similar, but why? A Toolkit for Explaining Text Similarity (2026.eacl-demo)
Copied to clipboard
| Challenge: | XPLAINSIM is a Python package that explains textual similarity in an easy-to-use way. |
| Approach: | They propose a Python package that unifies three approaches to explain text similarity . they demonstrate the value of the package through intuitive examples and empirical research . |
| Outcome: | XPLAINSIM is a Python package that unifies three approaches to explain text similarity . the authors show that the package is useful for explaining text similarities in a simple way . |
pyvene: A Library for Understanding and Improving PyTorch Models via Interventions (2024.naacl-demo)
Copied to clipboard
Zhengxuan Wu, Atticus Geiger, Aryaman Arora, Jing Huang, Zheng Wang, Noah Goodman, Christopher Manning, Christopher Potts
| Challenge: | Existing libraries are often project-based, but pyvene provides a unified and extensible framework for performing interventions on neural models and sharing the intervened upon models with others. |
| Approach: | They propose an open-source Python library that supports customizable interventions on a range of different PyTorch modules. |
| Outcome: | The proposed framework provides a unified and extensible framework for performing interventions on neural models and sharing the intervened upon models with others. |
Considering Likelihood in NLP Classification Explanations with Occlusion and Language Modeling (2020.acl-srw)
Copied to clipboard
| Challenge: | Existing explanation methods produce invalid or syntactically incorrect data, neglecting the improved abilities of recent NLP models. |
| Approach: | They propose an explanation method that combines occlusion and language models to sample valid and syntactically correct replacements with high likelihood, given the context of the original input. |
| Outcome: | The proposed method can sample valid and syntactically correct replacements with high likelihood, given the context of the original input. |
Interpretable Relevant Emotion Ranking with Event-Driven Attention (D19-1)
Copied to clipboard
| Challenge: | Existing studies ignore the latent event information in documents . Existing methods for detecting emotions are limited to a few words . |
| Approach: | They propose to integrate event information into a deep learning architecture to extract relevant emotion ranking models using corpus-level event embeddings and document-level events. |
| Outcome: | The proposed model performs better than state-of-the-art emotion detection and multi-label approaches on three real-world corpora and interpretable results shed light on the events which trigger certain emotions. |
BERT-Based Neural Collaborative Filtering and Fixed-Length Contiguous Tokens Explanation (2020.aacl-main)
Copied to clipboard
| Challenge: | Existing models that learn accurate representations of users and items are based on ratings, which oversimplify user preferences and item characteristics. |
| Approach: | They propose a novel, accurate, and explainable recommender model that integrates three key elements: BERT, multilayer perceptron, and maximum subarray problem to derive contextualized review features, model user-item interactions, and generate explanations. |
| Outcome: | The proposed model outperforms state-of-the-art models by an improvement gain of nearly 7% based on the human judges’ assessment . |
KERMIT: Complementing Transformer Architectures with Encoders of Explicit Syntactic Interpretations (2020.emnlp-main)
Copied to clipboard
Fabio Massimo Zanzotto, Andrea Santilli, Leonardo Ranaldi, Dario Onorati, Pierfrancesco Tommasino, Francesca Fallucchi
| Challenge: | Syntactic parsers are losing their centrality in downstream tasks due to the success of large-scale textual representation learners. |
| Approach: | They propose to embed symbolic syntactic parse trees into artificial neural networks to visualize how syntax is used in inference. |
| Outcome: | The proposed encoder can visualize how syntax is used in inference. |
Know-MRI: A Knowledge Mechanisms Revealer&Interpreter for Large Language Models (2025.acl-demo)
Copied to clipboard
Jiaxiang Liu, Boxuan Xing, Chenhao Yuan, ChenxiangZhang ChenxiangZhang, Di Wu, Xiusheng Huang, Haida Yu, Chuhan Lang, Pengfei Cao, Jun Zhao, Kang Liu
| Challenge: | Existing interpretation methods only support tasks with specific inputs, limiting their practical applications. |
| Approach: | They propose an extensible module that matches different input data with interpretation methods and consolidates the interpreting outputs. |
| Outcome: | The proposed module can match different input data with interpretation methods and consolidate the interpreting outputs. |
Improving Personalized Explanation Generation through Visualization (2022.acl-long)
Copied to clipboard
| Challenge: | Existing explainable recommendation models generate repetitive sentences for different items or empty sentences with insufficient details. |
| Approach: | They propose a visual-enhanced approach to generate rating scores and text explanations using visualization generation and text–image matching discrimination. |
| Outcome: | The proposed approach improves both the text quality and the diversity and explainability of the generated explanations. |
GREEN: Generative Radiology Report Evaluation and Error Notation (2024.findings-emnlp)
Copied to clipboard
Sophie Ostmeier, Justin Xu, Zhihong Chen, Maya Varma, Louis Blankemeier, Christian Bluethgen, Arne Md, Michael Moseley, Curtis Langlotz, Akshay Chaudhari, Jean-Benoit Delbrouck
| Challenge: | Existing automated evaluation metrics fail to consider factual correctness or are limited in their interpretability. |
| Approach: | They propose a radiology report evaluation metric that leverages natural language understanding of language models to identify and explain clinically significant errors. |
| Outcome: | The proposed method demonstrates higher correlation with expert error counts and higher alignment with expert preferences when compared to previous methods. |
How does Attention Affect the Model? (2021.findings-acl)
Copied to clipboard
| Challenge: | Existing studies on the effectiveness of attention in NLP do not consider changes in semantic capability of different components. |
| Approach: | They propose a framework that exploits a convex hull representation of sequence semantics in an n-dimensional Semantic Euclidean Space and defines indicators to capture the impact of attention on sequence semantic. |
| Outcome: | The proposed framework exploits a convex hull representation of sequence semantics in an n-dimensional Semantic Euclidean Space and defines indicators to capture the impact of attention on sequence semantic. |
XRec: Large Language Models for Explainable Recommendation (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Collaborative filtering (CF) is a widely adopted approach, but lacks the ability to provide explanations for the recommended items. |
| Approach: | They propose a model-agnostic framework that enables large language models to provide comprehensive explanations for user behaviors in recommender systems. |
| Outcome: | The proposed framework outperforms baseline approaches in explainable recommender systems. |
Decontextualization: Making Sentences Stand-Alone (2021.tacl-1)
Copied to clipboard
| Challenge: | Taking excerpts of text can be problematic, as key pieces may not be explicit in a local window. |
| Approach: | They define a problem of sentence decontextualization by rewriting a sentence to be interpretable out of context while preserving its meaning. |
| Outcome: | The proposed method can be used in question answering and document understanding tasks. |
PUCP-Metrix: An Open-source and Comprehensive Toolkit for Linguistic Analysis of Spanish Texts (2026.eacl-demo)
Copied to clipboard
| Challenge: | Existing tools for linguistic analysis of Spanish texts lack linguistic features for interpretability and tasks that involve style, structure, and readability. |
| Approach: | They propose to use PUCP-Metrix to analyze Spanish texts in a language repository. |
| Outcome: | The proposed toolkit performs better on automated readability assessments and machine-generated text detection tasks than existing repositories and strong neural baselines. |
Topic Modeling in Embedding Spaces (2020.tacl-1)
Copied to clipboard
| Challenge: | Existing topic models fail to learn interpretable topics when working with large and heavy-tailed vocabularies. |
| Approach: | They propose an embedded topic model that integrates word embeddings with a categorical distribution that is the natural parameter between the word’s embeddment and an embeddement of its assigned topic. |
| Outcome: | The embedded topic model outperforms existing topic models in terms of topic quality and predictive performance. |
ferret: a Framework for Benchmarking Explainers on Transformers (2023.eacl-demo)
Copied to clipboard
| Challenge: | Existing methods for interpreting transformer outputs are scattered and hard to operationalize. |
| Approach: | They propose a Python library to simplify the use and comparisons of XAI methods on transformers. |
| Outcome: | The proposed method provides better explanations and is preferable in the context of transformer models. |
VisDoT : Enhancing Visual Reasoning through Human-Like Interpretation Grounding and Decomposition of Thought (2026.findings-eacl)
Copied to clipboard
| Challenge: | Lack of perceptual grounding limits vision-language models' ability to interpret visual data . prior work on visualized data understanding focused on adapting VLMs to instruction tuning and chain-of-thought supervision . |
| Approach: | They propose a framework that enhances visual reasoning through human-like interpretation grounding. |
| Outcome: | The proposed framework improves on ChartQA and ChartQAPro benchmarks by +11.2%. |
Compositional Structured Explanation Generation with Dynamic Modularized Reasoning (2024.starsem-1)
Copied to clipboard
| Challenge: | Large-scale language models have shown remarkable performance on reasoning tasks such as reading comprehension, natural language inference, story generation, etc. |
| Approach: | They propose a compositional structured explanation generation task to test a model's ability to generalize from generating entailment trees to more steps, focusing on the length and shapes of engorgement trees. |
| Outcome: | The proposed model shows competitive compositional generalization abilities in a generation setting. |
Improve Interpretability of Neural Networks via Sparse Contrastive Coding (2022.findings-emnlp)
Copied to clipboard
| Challenge: | XAI has achieved remarkable advances, but few efforts have been devoted to solving the problem. |
| Approach: | They propose a model-agnostic explanation method termed Sparse Contrastive Coding . they use model-based explanations to explain the black-box in a more model-oriented way . |
| Outcome: | The proposed method outperforms five state-of-the-art methods in interpretability and classification metrics. |
CafGa: Customizing Feature Attributions to Explain Language Models (2025.emnlp-demos)
Copied to clipboard
| Challenge: | Feature attribution methods, such as SHAP and LIME, quantify the influence of each input component in a model. |
| Approach: | They propose a tool for generating and evaluating feature attribution explanations at customizable granularities. |
| Outcome: | The proposed tool is compared with two baseline methods: PartitionSHAP and MExGen. |
Clinical-Coder: Assigning Interpretable ICD-10 Codes to Chinese Clinical Notes (2020.acl-demos)
Copied to clipboard
| Challenge: | Existing methods of automatic coding prediction have been successful, but the interpretability of predicted codes is a challenge. |
| Approach: | They propose an online system that can predict ICD codes for Chinese clinical notes by using a Dilated Convolutional Attention network with N-gram Matching mechanism. |
| Outcome: | The proposed system is able to provide supporting information in clinical decision making. |
RAGVUE: A Diagnostic View for Explainable and Automated Evaluation of Retrieval-Augmented Generation (2026.eacl-demo)
Copied to clipboard
| Challenge: | Existing tools for evaluating RAG systems often collapse heterogeneous behaviors into single scores. |
| Approach: | They propose a diagnostic framework for automated, reference-free evaluation of RAG pipelines. |
| Outcome: | The proposed framework decomposes RAG behavior into retrieval quality,answer relevance and completeness, strictclaim-level faithfulness, and judge calibration. |
ELLA: Empowering LLMs for Interpretable, Accurate and Informative Legal Advice (2024.acl-demos)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have shown impressive performance in various tasks, showing great potential for specific domains, such as law (Lai et al., 2023), finance (Zeng e e al. 2023) and law (Lam elms, 2024). |
| Approach: | They propose to use large language models to provide interpretable, accurate, and informative legal advice by visually presenting the correlation between legal articles and LLM's response by calculating their similarities. |
| Outcome: | The proposed model provides users with an intuitive legal basis for the responses and retrieves relevant legal cases for user reference. |
Evaluating Readability and Faithfulness of Concept-based Explanations (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for evaluating concepts from different perspectives lack a unified formalization. |
| Approach: | They propose a formal definition of concepts generalizing to diverse concept-based explanations’ settings and apply it to other types of explanations or tasks. |
| Outcome: | Extensive experimental analysis was carried out to determine the evaluation measures for explanation evaluation measures. |
Explaining Why: How Instructions and User Interfaces Impact Annotator Rationales When Labeling Text Data (2022.naacl-main)
Copied to clipboard
Jamar Sullivan Jr., Will Brackenbury, Andrew McNutt, Kevin Bryson, Kwam Byll, Yuxin Chen, Michael Littman, Chenhao Tan, Blase Ur
| Challenge: | In the context of data labeling, researchers are interested in having humans select rationales . |
| Approach: | They conducted an online user study to understand how humans select rationales . they found that participants were near unanimous in their data labels . |
| Outcome: | The results show that participants selected 12% of input tokens as rationales, but fewer if unable to drag over multiple tokens at once. |
SFAL: Semantic-Functional Alignment Scores for Distributional Evaluation of Auto-Interpretability in Sparse Autoencoders (2025.emnlp-industry)
Copied to clipboard
| Challenge: | Interpreting the internal representations of large language models (LLMs) is crucial for their deployment in real-world applications, impacting areas such as AI safety, debugging, and compliance. |
| Approach: | They propose an alternative evaluation strategy that assesses the alignment between the semantic neighbourhoods of features and their functional neighbourhoods by using co-occurrence statistics. |
| Outcome: | The proposed evaluation strategy reduces reliance on scoring on large-scale models and improves efficiency and cost-effectiveness. |
AERA Chat: An Interactive Platform for Automated Explainable Student Answer Assessment (2025.emnlp-demos)
Copied to clipboard
| Challenge: | Existing systems that use pretrained language models to score student answers are noisy and unreliable. |
| Approach: | They propose a visualization platform for automated student answer assessment that leverages multiple LLMs to generate rationales. |
| Outcome: | The proposed platform enables educators to mark tasks and researchers to evaluate rationale quality from different models. |
Improving the Faithfulness of Attention-based Explanations with Task-specific Information for Text Classification (2021.acl-long)
Copied to clipboard
| Challenge: | Recent studies show that attention cannot be considered as a faithful explanation across encoders and tasks. |
| Approach: | They propose a new family of Task-Scaling mechanisms that scale attention weights across tasks and two attention mechanisms. |
| Outcome: | The proposed models improve explanation faithfulness across two attention mechanisms, five encoders and five text classification datasets without sacrificing predictive performance. |
A Differentiable Integer Linear Programming Solver for Explanation-Based Natural Language Inference (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing ILP frameworks are non-differentiable and cannot be integrated as part of a broader deep learning architecture. |
| Approach: | They propose a neuro-symbolic architecture for explanation-based NLI based on DBCS. |
| Outcome: | The proposed approach achieves superior performance when compared to existing solvers and black-box solver. |
From Generating Answers to Building Explanations: Integrating Multi-Round RAG and Causal Modeling for Scientific QA (2025.naacl-industry)
Copied to clipboard
Victor Barres, Clifton James McFate, Aditya Kalyanpur, Kailash Karthik Saravanakumar, Lori Moon, Natnael Seifu, Abraham Bautista-Castillo
| Challenge: | Application of Large Language Models to complex causal question answering can be stymied by their opacity and propensity for hallucination. |
| Approach: | They propose a causal QA approach that combines iterative RAG with a formal model of causation. |
| Outcome: | The proposed approach is implemented into a Collaborative Research Assistant (Cora) and evaluated in the life sciences domain. |
Semantic Accuracy in Natural Language Generation: A Thesis Proposal (2023.acl-srw)
Copied to clipboard
| Challenge: | Using large pre-trained language models, it is essential to research their reliability . if a human does not know the answer to a question, the socially acceptable behavior is to say 'I do not know' failing to fulfill this expectation can lead to distrust, or spread of misinformation. |
| Approach: | They propose a method for evaluating semantic accuracy and a benchmark for NLG metrics. |
| Outcome: | The proposed method evaluates semantic accuracy and provides a benchmark for NLG metrics. |
Aligning What LLMs Do and Say: Towards Self-Consistent Explanations (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are often prompted to produce natural language explanations, but the features driving the answer are often different from those emphasized in their explanations. |
| Approach: | They propose a large-scale benchmark linking model decisions with diverse explanations and attribution vectors across datasets, methods, and model families to address this gap. |
| Outcome: | The proposed model generates an answer where the word NLP in the prompt has high feature importance. |
Faithfulness Beyond Plausibility: Auditing Human Explanations in Educational Assessment (2026.acl-srw)
Copied to clipboard
| Challenge: | a gap exists between explanation components and how scores are constructed, and whether they reflect how scores were constructed . authors: explanation components are structurally inconsistent and may not be used as post-hoc justifications . |
| Approach: | They propose to use human tutor grading traces to test whether explanations are reliable . they find that removing rubric-level information leads to substantial changes in reconstructed scores . |
| Outcome: | The proposed diagnostic measures how explanation components contribute to score interpretation . removing rubric-level information leads to substantial changes in reconstructed scores . |
AbsInstruct: Eliciting Abstraction Ability from LLMs through Explanation Tuning with Plausibility Estimation (2024.acl-long)
Copied to clipboard
Zhaowei Wang, Wei Fan, Qing Zong, Hongming Zhang, Sehyun Choi, Tianqing Fang, Xin Liu, Yangqiu Song, Ginny Wong, Simon See
| Challenge: | Existing work shows that LLMs are deficient in abstract ability, and how to improve it remains unexplored. |
| Approach: | They propose a framework AbsInstruct to enhance LLMs’ abstract ability through instruction tuning. |
| Outcome: | The proposed framework can enhance LLMs’ abstraction ability with strong generalization performance while maintaining their general instruction-following abilities. |
TRACE: Training and Inference-Time Interpretability Analysis for Language Models (2025.emnlp-demos)
Copied to clipboard
| Challenge: | Existing tools for interpretability analysis of transformer models are post hoc, rely on scalar metrics or require nontrivial integration effort. |
| Approach: | They propose a modular toolkit for training and inference-time interpretability analysis of transformer models. |
| Outcome: | Experiments with autoregressive transformers show that TRACE reveals developmental phenomena overlooked by traditional scalar metrics such as loss or accuracy. |
SELFEXPLAIN: A Self-Explaining Architecture for Neural Text Classifiers (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing models that explain text classification predictions are opaque and overfit to spurious artifacts. |
| Approach: | They propose a novel self-explaining model that explains a text classifier’s predictions using phrase-based concepts. |
| Outcome: | The proposed model shows that it is adequate, trustworthy and understandable by human judges compared to existing baselines. |
Diff-Explainer: Differentiable Convex Optimization for Explainable Multi-hop Inference (2022.tacl-1)
Copied to clipboard
| Challenge: | Existing explainable multi-hop inference models are regarded as black-boxes due to their ability to transfer linguistic and semantic information to downstream tasks, posing concerns about interpretability and transparency of their predictions. |
| Approach: | They propose a hybrid framework that integrates explicit constraints with neural architectures through differentiable convex optimization to answer and explain multi-hop questions in natural language. |
| Outcome: | The proposed framework improves performance on scientific and commonsense QA tasks while still providing structured explanations in support of its predictions. |
Pragmatic Perspective on Assessing Implicit Meaning Interpretation in Sentiment Analysis Models (2025.acl-srw)
Copied to clipboard
| Challenge: | Using pragmatic theories of implicature, interpreting texts with implicit meaning correctly is essential for precise natural language understanding. |
| Approach: | They propose to use transformer models fine-tuned for sentiment analysis to illustrate the challenges in computational interpretation of implicatures. |
| Outcome: | The proposed model classifications reveal the limitations of supervised machine learning methods in detecting implicit sentiments. |
Explaining Language Model Predictions with High-Impact Concepts (2024.findings-eacl)
Copied to clipboard
| Challenge: | Existing methods to explain large language models (LLMs) are mostly correlational and lack causal features due to compositional nature of languages. |
| Approach: | They propose a framework to provide impact-aware explanations for large language models that are robust to feature changes and influential to the model’s predictions. |
| Outcome: | The proposed explanations improve on real and synthetic tasks and are robust to feature changes and influential to the model’s predictions. |
Protein2Text: Resampling Mechanism to Translate Protein Sequences into Human-Interpretable Text (2025.naacl-industry)
Copied to clipboard
Ala Jararweh, Oladimeji Macaulay, David Arredondo, Yue Hu, Luis E Tafoya, Kushal Virupakshappa, Avinash Sahu
| Challenge: | Proteins play critical roles in biological systems, yet 99.7% of 227 million known protein sequences remain uncharacterized due to the limitations of experimental methods. |
| Approach: | They propose a multimodal large language model that interprets protein sequences and generates informative text to address open-ended questions about protein functions and attributes. |
| Outcome: | The proposed model outperforms existing models in open-ended question-answering tasks. |
Rationalization through Concepts (2021.findings-acl)
Copied to clipboard
| Challenge: | Existing models that explain complex decisions are limited because of their lack of interpretability. |
| Approach: | They propose a model that extracts text snippets as concepts and infers which ones are described in the document. |
| Outcome: | The proposed model outperforms state-of-the-art methods trained on each aspect label independently. |
Actionable Interpretability for Churn Classification: A Text Bottleneck Model Case Study at a Major Telecom Provider (2026.acl-industry)
Copied to clipboard
| Challenge: | Managing customer churn is vital for subscription-based businesses . large language models (LLMs) can automate the classification of chursn-intent at scale . lack of transparency forces a difficult choice between automated systems and manual review . |
| Approach: | They propose to use text bottleneck models to classify customer churn in subscription-based businesses . they show that the model can be used to bridge the perceived trade-off between interpretability andpredictive performance . |
| Outcome: | The proposed model performs competitively with black-box baselines and an interactive dashboard. |
Towards Interpreting and Mitigating Shortcut Learning Behavior of NLU models (2021.naacl-main)
Copied to clipboard
Mengnan Du, Varun Manjunatha, Rajiv Jain, Ruchi Deshpande, Franck Dernoncourt, Jiuxiang Gu, Tong Sun, Xia Hu
| Challenge: | Recent studies indicate that NLU models are prone to rely on shortcut features for prediction, without achieving true language understanding. |
| Approach: | They propose a shortcut mitigation framework to suppress NLU models from making overconfident predictions for samples with large shortcut degree. |
| Outcome: | The proposed framework suppresses the model from making overconfident predictions for samples with large shortcut degree. |
From Wrong To Right: A Recursive Approach Towards Vision-Language Explanation (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for generating insightful explanations with limited annotations are limited. |
| Approach: | They propose a method that iteratively computes visual features, an answer, and an explanation to improve the explanation quality step by step until the answer converges. |
| Outcome: | The proposed method outperforms previous methods while utilizing 5% of the human-annotated explanations across 10 metrics, showing up to 4.2 and 1.3 increases in BLEU-1 score on the VCR and VQA-X datasets. |
Cross-Refine: Improving Natural Language Explanation Generation by Learning in Tandem (2025.coling-main)
Copied to clipboard
| Challenge: | Natural language explanations (NLEs) are vital for elucidating the reasoning behind large language model (LLM) decisions. |
| Approach: | They propose a role-modeling approach that employs two LLMs as generator and critic to generate and refine NLEs. |
| Outcome: | The proposed model outperforms self-refine and can perform with less powerful LLMs. |
Spectra: A Mechanistic Interpretability Library for Vision-Language Models (2026.acl-demo)
Copied to clipboard
| Challenge: | Existing interpretability tools for visionlanguage models are limited to activation probing and saving. |
| Approach: | They propose a library specifically designed for mechanistic interpretability of visionlanguage models that provides unified abstractions for activation patching, attention pattern analysis, and meta-functions across diverse VLM architectures. |
| Outcome: | The proposed library handles architecture-specific complexities while maintaining a simple, high-level interface. |
Locally Distributed Activation Vectors for Guided Feature Attribution (2022.coling-1)
Copied to clipboard
| Challenge: | Existing methods to explain predictions of deep neural networks are unstable and do not always provide faithful explanations to the target model. |
| Approach: | They propose a method to learn explanations-specific representations while constructing deep network models for text classification. |
| Outcome: | The proposed method improves model interpretability while preserving predictive performance. |
Identifying Linear Relational Concepts in Large Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | a technique for finding concept directions for human-interpretable concepts is needed to find their direction in the latent space . a linear relational concept (LRC) can be used to locate concepts in hidden activations . |
| Approach: | They propose a method for finding human-interpretable concepts by inverting a linear relational embedding and using earlier object layers. |
| Outcome: | The proposed method outperforms standard probing classifiers on performance as concept classifier and ability to causally change model output. |
Explaining Classes through Stable Word Attributions (2022.findings-acl)
Copied to clipboard
| Challenge: | Input saliency methods have become popular for explaining predictions of deep learning models, but there has been little work investigating methods for aggregating prediction-level explanations to the class level. |
| Approach: | They propose a method to aggregate prediction-level explanations to the class level using XLM-R and Integrated Gradients input attribution methods. |
| Outcome: | The proposed method extracts keyword lists of classes from text classification tasks and evaluates them on web register data. |
Normalized AOPC: Fixing Misleading Faithfulness Metrics for Feature Attributions Explainability (2025.acl-long)
Copied to clipboard
Joakim Edin, Andreas Geert Motzfeldt, Casper L. Christensen, Tuukka Ruotsalo, Lars Maaløe, Maria Maistro
| Challenge: | Deep neural network predictions are notoriously difficult to interpret due to the difficulty in understanding their inner mechanisms. |
| Approach: | They propose to normalize AOPC to enable consistent cross-model evaluations and more meaningful interpretation of individual scores. |
| Outcome: | The proposed approach can radically change AOPC results, questioning the conclusions of earlier studies and offering a more robust framework for assessing feature attribution faithfulness. |
E-VarM: Enhanced Variational Word Masks to Improve the Interpretability of Text Classification Models (2022.coling-1)
Copied to clipboard
Ling Ge, ChunMing Hu, Guanghui Ma, Junshuang Wu, Junfan Chen, JiHong Liu, Hong Zhang, Wenyi Qin, Richong Zhang
| Challenge: | Empirical studies show that our approach outperforms the SOTA methods in improving the interpretability of text classification models. |
| Approach: | They propose an enhanced variational word masks approach that exploits the Variational Information Bottleneck to obtain task-specific words. |
| Outcome: | Empirical results show that the proposed method outperforms the SOTA methods in improving the interpretability of the model. |
You May Like This Hotel Because ...: Identifying Evidence for Explainable Recommendations (2020.aacl-main)
Copied to clipboard
| Challenge: | Several studies have addressed explainable recommendations that produce natural language sentences . however, this task cannot explain detailed evidences for each hotel . |
| Approach: | They propose to decompose the process into two subtasks: Evidence Identification and Evidence Explanation. |
| Outcome: | The proposed model can explain evidences in recommending hotels given vague requests . it can find evidence sentences with respect to various vague requests and generate recommendation sentences . |
A Survey on Sparse Autoencoders: Interpreting the Internal Mechanisms of Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Sparse Autoencoders (SAEs) can disentangle complex features into more interpretable components. |
| Approach: | They propose to use Sparse Autoencoders to disentangle LLM features into more interpretable components. |
| Outcome: | The proposed method disentangles complex features into more interpretable components. |
Large Vision-Language Model Alignment and Misalignment: A Survey Through the Lens of Explainability (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Large Vision-Language Models have demonstrated remarkable capabilities in processing both visual and textual information. |
| Approach: | They examine the challenge of alignment and misalignment in LVLMs through an explainability lens. |
| Outcome: | The findings highlight the need for standardized evaluation protocols and in-depth explainability studies. |
Attention Consistency for LLMs Explanation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing interpretability methods face limitations such as low resolution and high computational cost. |
| Approach: | They propose a multi-layer attention consistency score to estimate the importance of input tokens in large language models. |
| Outcome: | The proposed heuristic achieves a favorable trade-off between interpretability quality and computational efficiency . |
Do Explanations Help Users Detect Errors in Open-Domain QA? An Evaluation of Spoken vs. Visual Explanations (2021.findings-acl)
Copied to clipboard
| Challenge: | despite interest in explainable AI, there is increasing skepticism as to whether explanations are useful to end-users in downstream applications. |
| Approach: | They conduct user studies to measure whether explanations help users decide when to accept or reject an ODQA system's answer. |
| Outcome: | The proposed study shows that explanations outperform baselines across modalities but the best strategy varies with the modality. |
Mechanistic Interpretability Should Prioritize Feature Consistency in Sparse Autoencoders (2026.acl-long)
Copied to clipboard
Xiangchen Song, Aashiq Muhamed, Yujia Zheng, Lingjing Kong, Zeyu Tang, Mona T. Diab, Virginia Smith, Kun Zhang
| Challenge: | Sparse Autoencoders (SAEs) are a tool in mechanistic interpretability (MI) but the aspiration to identify a canonical set of features is challenged by the observed inconsistency of learned SAE features across different training runs. |
| Approach: | They propose to use the Pairwise Dictionary Mean Correlation Coefficient to quantify SAE feature consistency as an evaluation axis alongside reconstruction and sparsity. |
| Outcome: | The proposed measure is based on the pairwise dictionary mean correlation coefficient (PW-MCC) on LLM activations. |
Explainable Prediction of Medical Codes from Clinical Text (N18-1)
Copied to clipboard
| Challenge: | Clinical notes are text documents that are created by clinicians for each patient encounter. |
| Approach: | They propose a method that aggregates information across the document using a convolutional neural network and uses an attention mechanism to select the most relevant segments for each of the thousands of possible codes. |
| Outcome: | The proposed method is accurate and better than the current state of the art. |
Value Profiles for Encoding Human Variation (2025.emnlp-main)
Copied to clipboard
Taylor Sorensen, Pushkar Mishra, Roma Patel, Michael Henry Tessler, Michiel A. Bakker, Georgina Evans, Iason Gabriel, Noah Goodman, Verena Rieser
| Challenge: | Using value profiles and a steerable decoder model to estimate ratings is crucial for personalization, pluralistic model alignment, and computational social science. |
| Approach: | They propose to represent individuals using value profiles and a steerable decoder model to estimate ratings conditioned on a value profile or other rater information. |
| Outcome: | The proposed model interpretably changes ratings according to semantic profile differences and is well-calibrated. |
Knowledge-Augmented Question Error Correction for Chinese Question Answer System with QuestionRAG (2025.emnlp-industry)
Copied to clipboard
| Challenge: | Large language models struggle with input errors, often failing to interpret user intent or altering the original question’s structure (over-correction). |
| Approach: | They propose a framework that uses reinforcement learning to address misinterpretation and over-correction by integrating external knowledge with the input. |
| Outcome: | The proposed framework unlocks the full potential of LLMs for the question correction task. |
EXPLAIN: Enhancing Retrieval-Augmented Generation with Entity Summary (2025.acl-industry)
Copied to clipboard
Yaozhen Liang, Xiao Liu, Jiajun Yu, Zhouhua Fang, Qunsheng Zou, Linghan Zheng, Yong Li, Zhiwei Liu, Haishuai Wang
| Challenge: | Existing document question answering methods reduce inference costs and input tokens. |
| Approach: | They propose a retrieval-augmented generation method that automatically extracts useful entities and generates summaries from documents. |
| Outcome: | The proposed method surpasses baseline retrieval-augmented generation (RAG) and long-context question answering (LC) methods achieve higher accuracy by processing entire documents, but at the cost of increased computational Corresponding authors. |
SAFR: Neuron Redistribution for Interpretability (2025.findings-naacl)
Copied to clipboard
| Challenge: | Existing studies on controlling neuron distribution for interpretability have focused on focusing on monosemanticity instead of focusing solely on feature interactions. |
| Approach: | They propose a method to regularize feature superposition by encoding representations of multiple features within a single neuron. |
| Outcome: | The proposed method improves model interpretability without compromising prediction performance. |
Word Embedding and WordNet Based Metaphor Identification and Interpretation (P18-1)
Copied to clipboard
| Challenge: | Existing models cannot identify exact metaphorical words within a sentence . current models do not rely on hand-crafted knowledge for training . |
| Approach: | They propose an unsupervised learning method that identifies and interprets metaphors at word-level without preprocessing. |
| Outcome: | The proposed method outperforms baseline models in two translation systems for English to Chinese showing that it paraphrases metaphors into their literal counterparts. |
Explainable Clinical Decision Support from Text (2020.emnlp-main)
Copied to clipboard
| Challenge: | Clinical prediction models often use structured variables and provide outcomes that are not readily interpretable by clinicians. |
| Approach: | They propose a hierarchical CNN-transformer model with explicit attention as an interpretable, multi-task clinical language model. |
| Outcome: | The proposed model achieves AUROCs of 0.75 and 0.78 on sepsis and mortality prediction. |
Learning Disentangled Semantic Spaces of Explanations via Invertible Neural Networks (2024.acl-long)
Copied to clipboard
| Challenge: | Disentangled latent spaces usually have better semantic separability and geometrical properties, which leads to better interpretability and controllable data generation. |
| Approach: | They propose a flow-based invertible neural network mechanism integrated with a transformer-based language Autoencoder to deliver latent spaces with better semantic separability and geometrical properties. |
| Outcome: | The proposed model can deliver latent spaces with better separability properties compared to the current state-of-the-art models. |
Contrastive Explanations for Model Interpretability (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for producing model explanations seek all causal factors at once, making them difficult to comprehend. |
| Approach: | They propose a method to produce contrastive explanations in the latent space . they use attribution and token/span attribution to produce models that consider only contrastive reasoning . |
| Outcome: | The proposed method allows model behavior to consider only contrastive reasoning . it also uncovers which aspects of the input are useful for and against particular decisions . |
Towards a Unified Model for Generating Answers and Explanations in Visual Question Answering (2023.findings-eacl)
Copied to clipboard
| Challenge: | Current explanation generation models are trained to select the best answers from Multiple-Choice questions or to classify single-word answers to a predetermined vocabulary. |
| Approach: | They propose a multitask learning approach towards a Unified Model for Answer and Explanation generation (UMAE) UMAE models surpass the prior state-of-the-art answer accuracy on A-OKVQA by 10 15%, show competitive results on OK-VQA and VCR, and demonstrate promising out-of domain performance on VQA-X. |
| Outcome: | The proposed model outperforms the state-of-the-art model on A-OKVQA and VCR and shows promising out-of domain performance on VQA-X. |
Towards Explaining Subjective Ground of Individuals on Social Media (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Large-scale language models have been reducing the gap between machines and humans in understanding how the real world works, yet understanding an individual’s perspective and behavior from text is far from being resolved. |
| Approach: | They propose a neural model that learns subjective grounds of individuals and accounts for their judgments on situations posted on social media. |
| Outcome: | The proposed model provides human-readable explanations of an individual’s subjective preference in judging social situations and also learns an individual's orientation towards abstract moral concepts. |
DoLFIn: Distributions over Latent Features for Interpretability (2020.coling-main)
Copied to clipboard
| Challenge: | Existing approaches to interpret neural networks face a trade-off between a model's usefulness and its complexity. |
| Approach: | They propose a novel approach to achieve interpretability that avoids this trade-off by using probability as the central quantity instead of a fixed quantity. |
| Outcome: | The proposed approach outperforms the classical CNN and BiLSTM classifiers on the SST2 and AG-news datasets. |
Do explanations make VQA models more predictable to a human? (D18-1)
Copied to clipboard
| Challenge: | Existing explanations of a model's behavior are not used in interactive tasks like Visual Question Answering (VQA). |
| Approach: | They analyze existing explanations and their role in making a VQA model more predictable to a human by using human-in-the-loop approaches that treat the model as a black-box. |
| Outcome: | The proposed explanations make a model more predictable to humans, whereas human-in-the-loop approaches treat it as a black-box do. |
ExplainCPE: A Free-text Explanation Benchmark of Chinese Pharmacist Examination (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing explanation datasets for large language models are limited to the English language and general domain, leading to a scarcity of linguistic diversity and a lack of resources in specialized domains, such as medical. |
| Approach: | They propose to use a medical dataset to assess the interpretability of Large Language Models (LLMs) . they propose to analyze medical text and generate rationales for their decisions . |
| Outcome: | The proposed model passes the pharmacist examination with a 75.7% accuracy, while other models like ChatGPT fail. |
Beyond Pointwise Scores: Decomposed Criteria-Based Evaluation of LLM Responses (2025.emnlp-industry)
Copied to clipboard
| Challenge: | DeCE is model-agnostic and domain-general, requiring no predefined taxonomies or handcrafted rubrics. |
| Approach: | They propose a decomposed LLM evaluation framework that separates accuracy and recall from accuracy and relevance. |
| Outcome: | The proposed framework achieves stronger correlation with expert judgments than traditional metrics and pointwise LLM scoring. |
Explaining Speech Classification Models via Word-Level Audio Segments and Paralinguistic Features (2024.eacl-long)
Copied to clipboard
| Challenge: | Existing explanations for speech classification models are difficult to interpret and make mistakes. |
| Approach: | They propose to explain speech classification models by using word-level and paralinguistic attributes to measure the impact of each audio segment aligned with a word on the outcome. |
| Outcome: | The proposed explanations correctly represent the model’s inner workings and are plausible to humans. |
Better Explain Transformers by Illuminating Important Information (2024.findings-eacl)
Copied to clipboard
| Challenge: | Existing explanations focus on the input and output of the Transformers, resulting in confusing results. |
| Approach: | They propose to highlight important information and eliminate irrelevant information by a refined information flow on top of the layer-wise relevance propagation method. |
| Outcome: | The proposed method outperforms baseline models on classification and question-answering datasets with over 3% to 33% improvement on explanation metrics. |
The Mechanical Bard: An Interpretable Machine Learning Approach to Shakespearean Sonnet Generation (2023.acl-short)
Copied to clipboard
| Challenge: | Rather than train a model to obey these constraints implicitly, we opt to enforce them explicitly using a simple but novel approach to generation. |
| Approach: | They propose to automate the generation of sonnets within preset poetic constraints using a constrained decoding approach that uses a relatively modest neural backbone. |
| Outcome: | The proposed method produces sonnets that adhere to the genre’s defined constraints and contain lyrical language and literary devices. |
Teaching Machine Comprehension with Compositional Explanations (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Recent advances in machine reading comprehension rely heavily on large-scale annotated corpora, which are timeconsuming and costly to collect. |
| Approach: | They propose to use semi-structured explanations to “teach” machines reading comprehension using a small number of semi-structural explanations that explicitly inform machines why answer spans are correct. |
| Outcome: | The proposed method achieves 70.14% F1 score with supervision from 26 explanations on the SQuAD dataset, comparable to plain supervised learning using 1,100 labeled instances yielding a 12x speed up. |
Generative Interpretation: Toward Human-Like Evaluation for Educational Question-Answer Pair Generation (2024.findings-eacl)
Copied to clipboard
| Challenge: | Existing evaluation methods often fail to produce objective results and favor high similarity to the ground-truth question-answer pairs. |
| Approach: | They propose an alternative approach to evaluate question-answer generation using Generative Interpretation (GI) GI outperforms existing evaluation methods in terms of human alignment . |
| Outcome: | The proposed approach outperforms existing evaluation methods in human alignment and shows comparable performance with GPT3.5, only with BART-large. |
Adjusting Interpretable Dimensions in Embedding Space with Human Judgments (2024.naacl-long)
Copied to clipboard
| Challenge: | Embedding spaces contain interpretable dimensions indicating gender, formality in style, or even object properties. |
| Approach: | They combine seed-based vectors with human ratings of where words fall along a specific dimension to evaluate on predicting object properties and stylistic properties. |
| Outcome: | The proposed model improves on seed-based vectors and human ratings on object properties and stylistic properties. |
Towards Explainable Computerized Adaptive Testing with Large Language Model (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods focus on minimizing the number of questions required to assess ability, lacking clear and reliable explanations for the question selection process. |
| Approach: | They propose to use large language models to enhance computer adaptive testing (CAT) by providing human-like interpretability and explanations. |
| Outcome: | The proposed agent-based CAT performs comparably or superior to traditional CAT methods in accuracy and significantly improves student trust and satisfaction. |
DecompX: Explaining Transformers Decisions by Propagating Token Decomposition (2023.acl-long)
Copied to clipboard
| Challenge: | Existing vector-based explanation methods for Transformer-based models are limited in their ability to explain the decisions of multiple layers. |
| Approach: | They propose a vector-based explanation method based on the construction of decomposed token representations and their successive propagation throughout the model without mixing them in between layers. |
| Outcome: | The proposed method outperforms existing vector-based and gradient-based methods on transformer-based models by a wide margin. |
Explaining Puzzle Solutions in Natural Language: An Exploratory Study on 6x6 Sudoku (2025.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are effective in human-AI collaborative decision-making, but they lack the ability to provide trustworthy, gradual, and tailored explanations. |
| Approach: | They evaluate the performance of five Large Language Models in solving and explaining Sudoku puzzles. |
| Outcome: | The proposed model can solve and explain complex Sudoku puzzles in a controlled environment. |
Retrieve to Explain: Evidence-driven Predictions for Explainable Drug Target Identification (2025.acl-long)
Copied to clipboard
| Challenge: | Existing language models lack the capability to quantitatively and faithfully compare answer plausibility in terms of supporting evidence. |
| Approach: | They propose a retrieval-based model that scores and ranks all possible answers to a research question based on evidence retrieved from a document corpus. |
| Outcome: | The proposed model scores and ranks all possible answers to a scientific question based on evidence retrieved from a document corpus . |
IrEne: Interpretable Energy Prediction for Transformers (2021.acl-long)
Copied to clipboard
| Challenge: | Existing software-based energy measurements of NLP models are not accurate because they do not consider the complex interactions between energy consumption and model execution. |
| Approach: | They propose an interpretable and extensible energy prediction system that predicts inference energy consumption of Transformer-based NLP models. |
| Outcome: | The proposed system predicts inference energy consumption of transformer models with an error of under 7% compared to the ground truth. |
Verification and Refinement of Natural Language Explanations through LLM-Symbolic Theorem Proving (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for assessing the validity of explanations for NLI are time-consuming and prone to logical errors. |
| Approach: | They propose a framework that integrates Large Language Models and Theorem Provers to verify and refine natural language explanations through crowd-sourcing . they propose to use TPs to generate and formalise explanatory sentences and suggest potential inference strategies for NLI. |
| Outcome: | The proposed framework generates and formalises explanatory sentences and suggests potential inference strategies for NLI. |
CodeExp: Explanatory Code Document Generation (2022.findings-emnlp)
Copied to clipboard
Haotian Cui, Chenglong Wang, Junjie Huang, Jeevana Priya Inala, Todd Mytkowicz, Bo Wang, Jianfeng Gao, Nan Duan
| Challenge: | Existing code-to-text generation models produce only high-level code summaries that do not capture implementation-level choices essential for these scenarios. |
| Approach: | They propose a code explanation generation task that uses code docstrings to refine models. |
| Outcome: | The proposed model can generate well-structured long docstrings comparable to human-written ones. |
Mitigating the Inconsistency Between Word Saliency and Model Confidence with Pathological Contrastive Training (2022.findings-acl)
Copied to clipboard
| Challenge: | Neural networks are used for various NLP tasks, but their complexity makes them difficult to interpret. |
| Approach: | They propose a framework to mitigate the model pathology and obtain more interpretable models by using contrastive learning and saliency-based samples augmentation to calibrate the sentences representation. |
| Outcome: | The proposed framework can mitigate the model pathology and generate more interpretable models while keeping the model performance. |
Interpretable Word Sense Representations via Definition Generation: The Case of Semantic Change Analysis (2023.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to semantic change analysis are limited in their interpretation power and lack of explanatory power. |
| Approach: | They propose to use specialised Flan-T5 language models to generate a definition for each usage and a specialised word sense model to generate the most prototypical definition. |
| Outcome: | The proposed representations outperform token or usage sentence embeddings in word-in-context semantic similarity judgements and are a promising type of lexical representation for NLP. |
FaithLM: Towards Faithful Explanations for Large Language Models (2026.eacl-long)
Copied to clipboard
Yu-Neng Chuang, Guanchu Wang, Chia-Yuan Chang, Ruixiang Tang, Shaochen Zhong, Fan Yang, Andrew Wen, Mengnan Du, Xuanting Cai, Vladimir Braverman, Xia Hu
| Challenge: | Large language models (LLMs) produce natural language explanations, but they lack faithfulness and do not reflect the evidence the model uses to decide. |
| Approach: | They propose a model-agnostic framework that evaluates and improves the faithfulness of LLM explanations without token masking or task-specific heuristics. |
| Outcome: | The proposed framework improves faithfulness of large language models without masking or heuristics. |
Comparing Explanation Faithfulness between Multilingual and Monolingual Fine-tuned Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Previous studies have investigated how different factors affect faithfulness of model explanations . |
| Approach: | They find that the larger the multilingual model, the less faithful FAs are compared to its counterpart monolingual models. |
| Outcome: | The results show that the larger the multilingual model, the less faithful the FAs are compared to its counterpart monolingual models. |
Paraphrases do not explain word analogies (2021.eacl-main)
Copied to clipboard
| Challenge: | Several attempts have been made to explain distributional word embeddings as linguistic regularities as directions. |
| Approach: | They propose to use an analogy to explain why linguistic regularities should hold in distributional word embeddings. |
| Outcome: | The proposed explanation does not hold empirically. |
The Solvability of Interpretability Evaluation Metrics (2023.findings-eacl)
Copied to clipboard
| Challenge: | Feature attribution methods are often evaluated on metrics such as comprehensiveness and sufficiency. |
| Approach: | They propose to use beam search to define problem of optimizing an explanation for a metric . they also propose to evaluate the metric on one or more metrics to determine its solvability . |
| Outcome: | The proposed explainer can solve the problem of optimizing an explanation for a metric by beam search. |
Beyond Coherence: Improving Temporal Consistency and Interpretability in Dynamic Topic Models (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing topic models capture bag-of-words statistics but lack semantic priors . interpretability remains shallow, relying on noisy top-word lists that obscure thematic clarity. |
| Approach: | They propose a variational framework to capture more faithful temporal trajectories . they propose to use entropy-regularized optimal transport to align entire topic constellations . |
| Outcome: | The proposed framework captures more faithful temporal trajectories and improves interpretability. |
GEAR: A Scalable and Interpretable Evaluation Framework for RAG-Based Car Assistant Systems (2025.emnlp-industry)
Copied to clipboard
Niloufar Beyranvand, Hamidreza Dastmalchi, Aijun An, Heidar Davoudi, Winston Chan, Ron DiCarlantonio
| Challenge: | Large language models (LLMs) increasingly power car assistants, but evaluating response quality remains a challenge. |
| Approach: | They propose a framework that uses large language models as evaluators to compare assistant responses against ground-truth counterparts. |
| Outcome: | The proposed framework compares assistant responses against ground-truth counterparts, assessing coverage, correctness, and other dimensions of answer quality. |
Scoring and Classifying Implicit Positive Interpretations: A Challenge of Class Imbalance (C18-1)
Copied to clipboard
| Challenge: | a reimplementation of a system on detecting implicit positive meaning from negated statements is reported . a baseline taking the mean score or most frequent class is hard to beat because of class imbalance in the dataset. |
| Approach: | They propose a system to detect implicit positive meaning from negated statements . they convert the scores into classes and report their results on regression and classification tasks . |
| Outcome: | The proposed system is hard to beat because of class imbalance in the dataset. |
Towards Understanding Text Factors in Oral Reading (N18-1)
Copied to clipboard
| Challenge: | Using a case study, we show that variation in oral reading rate is consistent across readers. |
| Approach: | They propose to use text complexity to predict reading rate for professional narrators . they also show that variation can be explained by timing and story-based factors . |
| Outcome: | The authors show that variation in reading rate can be explained by features of the texts being read. |
KACE: Generating Knowledge Aware Contrastive Explanations for Natural Language Inference (2021.acl-long)
Copied to clipboard
| Challenge: | Existing approaches in NLP focus on “WHY A” rather than contrastive “WHA NOT B” Experimental results show that contrastive explanations are beneficial to fit the scenarios by clarifying the difference between the predicted answer and other possible wrong ones. |
| Approach: | They propose to generate contrastive explanations with counterfactual examples in NLI by identifying key phrases from input sentences and using them as key perturbations to generate counterfacts. |
| Outcome: | The proposed framework improves on SNLI and ETPA models by 91.9%. |
A Study of Automatic Metrics for the Evaluation of Natural Language Explanations (2021.eacl-main)
Copied to clipboard
| Challenge: | a lack of transparency is a key issue for robotics and AI. |
| Approach: | They propose to map existing automatic evaluation methods for natural language generation onto explanations. |
| Outcome: | The proposed model shows that embedding-based evaluation methods have higher correlations with human ratings than word-overlap metrics. |
Automating Legal Interpretation with LLMs: Retrieval, Generation, and Evaluation (2025.acl-long)
Copied to clipboard
| Challenge: | a novel framework for automated legal interpretation is proposed to alleviate the burden on legal experts. |
| Approach: | They propose a framework for automated legal interpretation that uses large language models to extract concept-related information and interpret legal concepts. |
| Outcome: | The proposed framework eliminates the need for legal experts to interpret legal concepts . it uses large language models to extract concept-related information and interpret legal concept interpretations . |
StyLEx: Explaining Style Using Human Lexical Annotations (2023.eacl-main)
Copied to clipboard
| Challenge: | Large pre-trained language models often learn spurious domain-specific words to make predictions. |
| Approach: | They propose a model that learns from human annotated explanations of stylistic features and jointly predicts them as model explanations. |
| Outcome: | The proposed model can provide human like stylistic lexical explanations without sacrificing performance on in-domain and out-of-domain datasets. |
SummaCoz: A Dataset for Improving the Interpretability of Factual Consistency Detection for Summarization (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Summarization is an important application of Large Language Models. |
| Approach: | They integrate human-annotated and model-generated natural language explanations to elucidate how a summary deviates and becomes inconsistent with its source article. |
| Outcome: | The proposed model provides rationales for its judgments and improves its accuracy significantly. |
Interpretable Safety Alignment via SAE-Constructed Low-Rank Subspace Adaptation (2026.acl-long)
Copied to clipboard
| Challenge: | Prior work has shown that safety behaviors are governed by low-rank structures . Low-Rank Adaptation (LoRA) consistently underperforms full fine-tuning and reinforcement learning on safety benchmarks . |
| Approach: | They propose a safety alignment system that disentangles safety-relevant directions into monosemantic features and constructs an interpretable safety subspace from SAE directions. |
| Outcome: | Empirically, the proposed model achieves 99.6% safety rates across multiple model families and scales . low-rank Adaptation consistently underperforms full fine-tuning and reinforcement learning on safety benchmarks compared with previous methods . |
A Joint Learning Framework for Restaurant Survival Prediction and Explanation (2022.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in deep learning have various models that research reviews and interactions for different kinds of tasks, such as predicting restaurant survival. |
| Approach: | They propose a joint learning framework for explainable restaurant survival prediction based on multi-modal data of user-restaurant interactions and users’ textual reviews. |
| Outcome: | The proposed framework improves on two datasets showing that it can model restaurant interactions and users’ textual reviews. |
ProtoLens: Advancing Prototype Learning for Fine-Grained Interpretability in Text Classification (2025.acl-long)
Copied to clipboard
| Challenge: | ProtoLens provides fine-grained, sub-sentence level interpretability for text classification. |
| Approach: | They propose a prototype-based model that provides fine-grained, sub-sentence level interpretability for text classification. |
| Outcome: | Extensive experiments show that ProtoLens outperforms both prototype-based and non-interpretable baselines on multiple text classification benchmarks. |
Lifelong Explainer for Lifelong Learners (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing explanation methods are inefficient when explaining a static black-box model. |
| Approach: | They propose a Lifelong Explanation approach that continuously trains a student explainer under the supervision of a teacher on different tasks undertaken in LL. |
| Outcome: | The proposed approach can be extended to include a teacher and maintain the same level of faithfulness to the black-box model as the student explainer while being up to 102 times faster at test time. |
Towards Reinterpreting Neural Topic Models via Composite Activations (2022.emnlp-main)
Copied to clipboard
| Challenge: | Most Neural Topic Models (NTMs) use a variational auto-encoder framework producing K topics limited to the size of the encoder’s output. |
| Approach: | They propose a model-free two-stage process to reinterpret NTM and derive further insights on the state of the trained model. |
| Outcome: | The proposed model-free process decouples the strict interpretation of topics from the original NTM and evaluates them on a large external corpus. |
Faithfully Explainable Recommendation via Neural Logic Reasoning (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing models for explainable recommendation have neglected faithfulness of KG reasoning . |
| Approach: | They propose to draw on interpretable logical rules to guide path-reasoning process for explanation generation. |
| Outcome: | The proposed method delivers high-quality recommendations and ascertains the faithfulness of the derived explanation. |
Analyzing (In)Abilities of SAEs via Formal Languages (2025.naacl-long)
Copied to clipboard
| Challenge: | Autoencoders have been used for finding interpretable and disentangled features underlying neural network representations in both image and text domains, but there is a lack of corresponding results for the text domain. |
| Approach: | They propose to train sparse autoencoders (SAEs) on a synthetic testbed of formal languages to find interpretable latents in models trained on formal languages. |
| Outcome: | The proposed approach promotes learning of causally relevant features in a formal language setting. |
Legal Judgment Reimagined: PredEx and the Rise of Intelligent AI Interpretation in Indian Courts (2024.findings-acl)
Copied to clipboard
| Challenge: | Prediction with Explanation is the largest expert-annotated dataset for legal judgment prediction and explanation in the Indian context . |
| Approach: | They propose to use an annotated legal judgment prediction corpus to improve models' accuracy . they employ transformer-based models tailored for both general and Indian legal contexts . |
| Outcome: | The proposed system improves the accuracy and explanatory depth of models for legal judgments. |
RAG-KT: Cross-platform Explainable Knowledge Tracing with Multi-view Fusion Retrieval Generation (2026.findings-acl)
Copied to clipboard
| Challenge: | Conventional Deep Learning (DL)-based KT models are tied to platform-specific identifiers and latent representations, making them hard to transfer and interpret. |
| Approach: | They propose a retrieval-augmented paradigm that frames cross-platform KT as reliable context constrained inference with LLMs. |
| Outcome: | Experiments on three public KT benchmarks show that the proposed paradigm improves accuracy and robustness, and also shows strong performance under cross-platform conditions. |
Proto-lm: A Prototypical Network-Based Framework for Built-in Interpretability in Large Language Models (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for interpreting LLMs are post hoc and focus on low-level features and lack of explainability at higher-level text units. |
| Approach: | They propose a prototypical network-based white-box framework that allows LLMs to learn immediately interpretable embeddings during the fine-tuning stage while maintaining competitive performance. |
| Outcome: | The proposed framework can learn interpretable embeddings during the fine-tuning stage while maintaining competitive performance. |
Exploring the Trade-off Between Model Performance and Explanation Plausibility of Text Classifiers Using Human Rationales (2024.findings-naacl)
Copied to clipboard
| Challenge: | Saliency post-hoc explainability methods are important tools for understanding complex NLP models, but they may not align with human intuition, making the explanations not plausible. |
| Approach: | They propose a method for incorporating rationales into text classification models by augmenting the standard cross-entropy loss with a novel loss function inspired by contrastive learning. |
| Outcome: | The proposed approach enhances the plausibility of post-hoc explanations while preserving their faithfulness. |
A Diagnostic Study of Explainability Techniques for Text Classification (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing explainability techniques that can be produced post-hoc with already trained models are lacking a definitive guide on how to choose one given a particular task and model architecture. |
| Approach: | They propose to use a list of diagnostic properties to evaluate existing explainability techniques to compare them with human annotations of salient input regions. |
| Outcome: | The proposed list compares a set of explainability techniques on downstream text classification tasks and neural network architectures. |
SenteCon: Leveraging Lexicons to Learn Human-Interpretable Language Representations (2023.findings-acl)
Copied to clipboard
| Challenge: | In many settings, it is important to understand a model’s decision-making process. |
| Approach: | They propose a method for introducing human interpretability in deep language representations by encoding a passage of text as a layer of interpretable categories. |
| Outcome: | The proposed method outperforms existing interpretable language representations on downstream tasks and on agreement with human characterizations of the text. |
SyntaxShap: Syntax-aware Explainability Method for Text Generation (2024.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for explaining sequence-to-sequence tasks are not suitable for textual data. |
| Approach: | They propose a local, model-agnostic explainability method that takes into account the syntax in text data and extends Shapley values to account for parsing-based syntactic dependencies. |
| Outcome: | The proposed method is compared to state-of-the-art explainability methods for text generation tasks using various metrics including faithfulness, coherency, and semantic alignment of the explanations to the model. |
Diagnosis of Dysarthria Severity and Explanation Generation Using XAI-Enhanced CLINIC-GENIE on Diadochokinetic Tasks (2026.findings-eacl)
Copied to clipboard
| Challenge: | Recent deep learning approaches for dysarthria impairment severity lack interpretability essential for clinical applications. |
| Approach: | They propose a deep neural network classifier that integrates acoustic and speech embeddings with Clinically Explainable Acoustic Features (CEAFs) and a module that transforms CEAFs and their Shapley values into intuitive natural language explanations. |
| Outcome: | The proposed model achieves a balanced accuracy of 0.952 (17.3% improvement over using CEAFs alone) and certified speech-language pathologists rated explanations with an average fidelity score of 4.94, confirming enhanced clinical utility. |
Decomposing Co-occurrence Matrices into Interpretable Components as Formal Concepts (2024.findings-acl)
Copied to clipboard
| Challenge: | Existing studies have shown that word vectors capture relational meanings, but the interpretability of their dimensions remains an open issue. |
| Approach: | They employ the mathematical methodology of Formal Concept Analysis to examine word embeddings using a count-based co-occurrence matrix. |
| Outcome: | The proposed model shows that the formal concepts identified align with interpretable categories, as shown in the category completion task. |
Is Attention Interpretable? (P19-1)
Copied to clipboard
| Challenge: | Attention mechanisms have recently boosted performance on a range of NLP tasks. |
| Approach: | They propose to manipulate attention weights in text classification models and analyze the resulting differences in their predictions. |
| Outcome: | The proposed approach improves models' predictions by using gradient-based rankings of attention weights. |
Neuro-Symbolic Reinforcement Learning with First-Order Logic (2021.emnlp-main)
Copied to clipboard
Daiki Kimura, Masaki Ono, Subhajit Chaudhury, Ryosuke Kohita, Akifumi Wachi, Don Joven Agravante, Michiaki Tatsubori, Asim Munawar, Alexander Gray
| Challenge: | Existing deep reinforcement learning methods require many trials before convergence and no direct interpretability of trained policies is provided. |
| Approach: | They propose a novel RL method which can learn symbolic and interpretable rules in their differentiable network. |
| Outcome: | The proposed method can learn symbolic and interpretable rules in their differentiable network. |
Transparency at the Source: Evaluating and Interpreting Language Models With Access to the True Distribution (2023.findings-emnlp)
Copied to clipboard
| Challenge: | a new approach to train, evaluate and interpret neural language models uses artificial, language-like data. |
| Approach: | They propose a setup for training, evaluating and interpreting neural language models that uses artificial, language-like data. |
| Outcome: | The proposed model is based on a massive probabilistic grammar and a large natural language corpus, and provides complete control over the generative process. |
What Gets Echoed? Understanding the “Pointers” in Explanations of Persuasive Arguments (D19-1)
Copied to clipboard
| Challenge: | Explanations are central to everyday life, and are a topic of growing interest in the AI community. |
| Approach: | They propose a word-level prediction task to investigate how explanations selectively reuse information from what is being explained. |
| Outcome: | The proposed features have strong predictive power on the echoing of a word in an explanation, and enhance neural methods of generating explanations. |
Faithful Serum: Mitigating the Faithfulness Gap in Textual Explanations of LLM Decisions via Attribution Guidance (2026.acl-long)
Copied to clipboard
| Challenge: | Prior work has focused on generating convincing rationales that appear to be subjectively faithful, but it remains unclear whether these explanations are epistemic faithful. |
| Approach: | They propose a method that enhances epistemic faithfulness by guiding explanation generation through attention-level interventions, informed by token-level heatmaps. |
| Outcome: | The proposed method significantly improves epistemic faithfulness across multiple models, benchmarks, and prompts. |
Topic Model or Topic Twaddle? Re-evaluating Semantic Interpretability Measures (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing methods for topic model evaluation use automated measures modeled on human evaluation tests that are dissimilar to applied usage. |
| Approach: | They propose to use a novel experimental framework to evaluate topic models and assess their coherence for specialized collections in an applied setting. |
| Outcome: | The proposed framework is reflective of human evaluations using open labeling, typical of applied research. |
What-if I ask you to explain: Explaining the effects of perturbations in procedural text (2020.findings-emnlp)
Copied to clipboard
| Challenge: | QUARTET constructs explanations from paragraphs using procedural text . qartet achieves 18 points better on explanation accuracy compared to strong baselines on a recent process comprehension benchmark. |
| Approach: | They propose a system that constructs explanations from paragraphs by modeling the explanation task as a multitask learning problem. |
| Outcome: | The proposed system achieves 18 points better on explanation accuracy compared to strong baselines on a process comprehension benchmark. |
Prototype-Based Interpretability for Legal Citation Prediction (2023.findings-acl)
Copied to clipboard
| Challenge: | citation prediction is a key problem in high-stakes decision making areas such as law . experts often require interpretability for automatic systems to be utilized in practical settings . |
| Approach: | They propose to use legal citation prediction to solve a problem with legal experts' feedback . they propose to add a prototype architecture to add interpretability while adhering to legal parameters . |
| Outcome: | The proposed model performs well while adhering to decision parameters used by lawyers. |
ExPUNations: Augmenting Puns with Keywords and Explanations (2022.emnlp-main)
Copied to clipboard
Jiao Sun, Anjali Narayan-Chen, Shereen Oraby, Alessandra Cervone, Tagyoung Chung, Jing Huang, Yang Liu, Nanyun Peng
| Challenge: | Puns add the challenge of fusing commonsense and world knowledge with the ability to interpret lexical-semantic ambiguity. |
| Approach: | They propose to augment existing datasets with detailed crowdsourced annotations of puns, keywords and fine-grained funniness ratings to challenge current models' ability to understand and generate humor. |
| Outcome: | The proposed tasks include explanation generation to aid with pun classification and keyword-conditioned pun generation to challenge state-of-the-art models' ability to understand and generate humor. |
Biomedical Interpretable Entity Representations (2021.findings-acl)
Copied to clipboard
| Challenge: | Existing work on general interpretable representation learning does not transfer to biomedicine . pre-trained models induce dense entity representations but are not immediately interpretable. |
| Approach: | They propose a method that exploits BIER's final sparse and intermediate dense representations to facilitate model and entity type debugging. |
| Outcome: | The proposed model performs well on biomedical tasks including disambiguation and label classification. |
Understanding Attention for Text Classification (2020.acl-main)
Copied to clipboard
| Challenge: | Existing studies have focused on whether local attention weights reflect the importance of input representations. |
| Approach: | They propose to analyze for each word token the following two quantities: its polarity score and its attention score, where the latter is a global assessment on the token’s significance. |
| Outcome: | The proposed model can be improved under conditions where the interplay between the two quantities can contribute towards model performance. |
COCKATIEL: COntinuous Concept ranKed ATtribution with Interpretable ELements for explaining neural net classifiers on NLP (2023.findings-acl)
Copied to clipboard
| Challenge: | Recent debates have shown that attention maps and attribution methods are unreliable. |
| Approach: | They propose a model-agnostic XAI technique that generates meaningful explanations from the last layer of a neural net model trained on an NLP classification task by using Non-Negative Matrix Factorization to discover concepts the model leverages to make predictions. |
| Outcome: | The proposed technique generates meaningful explanations from the last layer of a neural net model trained on an NLP classification task without compromising the accuracy of the underlying model or requiring a new one to be trained. |
Comparative Analysis of Sign Language Interpreting Agents Perception: A Study of the Deaf (2024.lrec-main)
Copied to clipboard
| Challenge: | Prior research on sign language recognition has shown encouraging outcomes in achieving highly accurate and dependable automatic sign language generation. |
| Approach: | They propose to compare a state-of-the-art sign language generation system with a skilled sign language interpreter to gain insights into usability of such metrics for deaf signers. |
| Outcome: | The proposed system is compared with a skilled interpreter to gain insights into usability of such metrics for deaf signers and how deafic signers perceive signing agents. |
Bayesian Compression for Natural Language Processing (D18-1)
Copied to clipboard
| Challenge: | In natural language processing, recurrent neural networks have a huge number of parameters. |
| Approach: | They propose a Bayesian sparsification technique which allows compressing RNNs dozens or hundreds of times without time-consuming hyperparameters tuning. |
| Outcome: | The proposed technique compresses the RNN dozens or hundreds of times without time-consuming hyperparameters tuning. |
Discovering Semantic Subdimensions through Disentangled Conceptual Representations (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches focus on predefined dimensions that overlook finer conceptual distinctions . a new framework is proposed to investigate the subdimensions underlying coarse-grained semantic dimensions . |
| Approach: | They propose a framework that decomposes word embeddings into multiple sub-embeddings . they propose to map these subdimensions to brain activation to assess their plausibility . |
| Outcome: | The proposed framework decomposes word embeddings from large language models into sub-embeddings, each encoding specific semantic information. |
Identification of Multiple Logical Interpretations in Counter-Arguments (2025.emnlp-main)
Copied to clipboard
Wenzhi Wang, Paul Reisert, Shoichi Naito, Naoya Inoue, Machi Shimmei, Surawat Pothong, Jungmin Choi, Kentaro Inui
| Challenge: | Counter-arguments (CAs) are a good way to improve learners' critical thinking skills . however, it is difficult to provide every learner tailored feedback due to limited human resources and heavy workloads. |
| Approach: | They propose to annotate a dataset of 134 CAs annotated with 13 logical predicate questions and train a model with Reinforcement Learning with Verifiable Rewards to identify multiple logical interpretations. |
| Outcome: | The proposed model performs on par with larger proprietary models. |
On Measuring Faithfulness or Self-consistency of Natural Language Explanations (2024.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) can explain their predictions through post-hoc or Chain-of-Thought explanations, but they can also make unfaithful explanations that hide their sensitivity to biasing inputs. |
| Approach: | They propose to use a model-based consistency test to judge the faithfulness of post-hoc or Chain-of-Thought explanations rather than model-specific faithfulness tests. |
| Outcome: | The proposed tests do not measure faithfulness to model explainability but rather their self-consistency at output level. |
CoNLL#: Fine-grained Error Analysis and a Corrected Test Set for CoNLL-03 English (2024.lrec-main)
Copied to clipboard
| Challenge: | a glass ceiling for named entity recognition systems has been suggested for 2021 . however, the performance of the most popular NER benchmarks has plateaued since then . we investigate what NER models are still struggling with . |
| Approach: | They perform a fine-grained evaluation of the model outputs by adding document annotations to the CoNLL-03 English dataset to identify lingering errors. |
| Outcome: | The proposed model is able to correct errors and guide future work. |
Learning to Judge: LLMs Designing and Applying Evaluation Rubrics (2026.findings-eacl)
Copied to clipboard
| Challenge: | Large language models are increasingly used as evaluators for natural language generation . human rubrics are often static and misaligned with how models internally represent language quality. |
| Approach: | They propose to use large language models to generate interpretable and task-aware evaluation dimensions and apply them within models. |
| Outcome: | The proposed model improves the semantic coherence and scoring reliability of LLM-defined criteria and their alignment with human criteria. |
Learning Explainable Linguistic Expressions with Neural Inductive Logic Programming for Sentence Classification (2020.emnlp-main)
Copied to clipboard
Prithviraj Sen, Marina Danilevsky, Yunyao Li, Siddhartha Brahma, Matthias Boehm, Laura Chiticariu, Rajasekar Krishnamurthy
| Challenge: | Existing approaches to explain models are difficult to interpret and have undesirable biases. |
| Approach: | They propose a neural network architecture for learning transparent sentences . they use linguistic expressions built on top of predicates extracted using shallow natural language understanding . |
| Outcome: | The proposed model outperforms statistical relational learning and other neuro-symbolic methods and performs better than black-box recurrent neural networks. |
ClozEx: A Task toward Generation of English Cloze Explanation (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing tasks and datasets specifically designed for generating language learner explanations for cloze questions are lacking . clozing questions are used to assess language proficiency and enhance language learning . |
| Approach: | They propose a task ClozEx to generate explanations for cloze questions in LA . they use a curated dataset of clozing questions paired with explanations . |
| Outcome: | The proposed task generates fluent explanations for cloze questions in English as a second language learners. |
Learning Variational Word Masks to Improve the Interpretability of Neural Text Classifiers (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for improving model interpretability require prior information or human annotations as additional inputs. |
| Approach: | They propose a variational word mask method to automatically learn task-specific important words and reduce irrelevant information on classification, which ultimately improves model interpretability. |
| Outcome: | The proposed method improves model prediction accuracy and interpretability on seven datasets. |
What Factors Should Paper-Reviewer Assignments Rely On? Community Perspectives on Issues and Ideals in Conference Peer-Review (2022.naacl-main)
Copied to clipboard
| Challenge: | a survey of the NLP community shows that paper-reviewer matching is a problem . authors lose valuable time and opportunities by writing reviews that are arbitrarily low . |
| Approach: | They propose to use paper-reviewer matching to improve peer review . they identify common issues and perspectives on what factors should be considered . |
| Outcome: | The proposed method improves the quality of peer review and improves interpretable peer review assignments. |
Attention is not Explanation (N19-1)
Copied to clipboard
| Challenge: | Attention mechanisms have seen wide adoption in neural NLP models. |
| Approach: | They perform extensive experiments to assess the degree to which attention weights provide meaningful "explanations" they find that attention weighted inputs are often uncorrelated with gradient-based measures of feature importance . |
| Outcome: | The proposed model is based on a distribution over attended-to input units . the findings show that attention weights are often uncorrelated with features . |
Ranking and Selecting Multi-Hop Knowledge Paths to Better Predict Human Needs (N19-1)
Copied to clipboard
| Challenge: | Existing methods to categorize sentiments and emotions in text are limited. |
| Approach: | They propose to extract, rank, filter and select multi-hop relation paths from a commonsense knowledge resource to interpret the expression of sentiment in terms of their underlying human needs. |
| Outcome: | The proposed method improves on a recently published dataset for categorizing human needs. |
GLUCOSE: GeneraLized and COntextualized Story Explanations (2020.emnlp-main)
Copied to clipboard
Nasrin Mostafazadeh, Aditya Kalyanpur, Lori Moon, David Buchanan, Lauren Berkowitz, Or Biran, Jennifer Chu-Carroll
| Challenge: | Existing knowledge resources and pretrained language models do not include or readily predict GLUCOSE’s rich inferential content. |
| Approach: | They propose a platform for crowdsourcing GLUCOSE data at scale that uses semi-structured templates to elicit causal explanations. |
| Outcome: | The proposed model can be trained on human-readable stories and build similar models on unseen stories. |
Causal Explanation Analysis on Social Media (D18-1)
Copied to clipboard
| Challenge: | Understanding causal explanations is an important psychological factor linked to physical and mental health. |
| Approach: | They propose to automate causal explanation analysis by building on discourse parsing and using a hierarchy of Bidirectional LSTMs to identify the specific phrase that is the explanation. |
| Outcome: | The proposed subtasks achieve strong accuracies but differ in their approaches . the proposed sub task is compared with the previous task and is able to identify the specific phrase that is the explanation. |
Assessing Digital Language Support on a Global Scale (2022.coling-1)
Copied to clipboard
| Challenge: | a new method is being developed to assess how well each language is doing in terms of digital language support. |
| Approach: | They develop an automated method to assess how well each language is doing in terms of digital language support. |
| Outcome: | The proposed method scrapes the names of supported languages from 143 digital tools and produces an explainable model for quantifying and monitoring it on a global scale. |
An Interpretable Neural Network with Topical Information for Relevant Emotion Ranking (D18-1)
Copied to clipboard
| Challenge: | Experimental results show that the extracted emotion-associated topic words represent emotion-evoking events. |
| Approach: | They propose an interpretable neural network approach for relevant emotion ranking . they initialize the hidden layer to approximate the behavior of topic models . |
| Outcome: | The proposed approach performs better than state-of-the-art methods on real-world corpora. |
ExaGPT: Example-Based Machine-Generated Text Detection for Human Interpretability (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing interpretable detectors are not aligned with the human decision-making process and fail to offer evidence that users easily understand. |
| Approach: | They propose an interpretable detection approach that checks whether a text is human-written or LLM-generated by checking whether it shares more similar spans with human-generated texts. |
| Outcome: | ExaGPT outperforms interpretable detectors by +37.0 points at a false positive rate of 1%. |
Personalized Transformer for Explainable Recommendation (2021.acl-long)
Copied to clipboard
| Challenge: | Recent years have witnessed the successful application of natural language generation. |
| Approach: | They propose a model that uses user and item IDs to predict the words in the target explanation to make personalized Transformer. |
| Outcome: | The proposed model outperforms BERT on the explainable recommendation task in terms of effectiveness and efficiency. |
PsyScore: A Psychometrically-Aware Framework for Trait-Adaptive Essay Scoring and ZPD-Scaffolded Feedback (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to Automated Essay Scoring (AES) treat scoring and feedback as separate components, resulting in fragmentation. |
| Approach: | They propose a psychometrically-aware framework that integrates diagnostic assessment with instructional scaffolding through a shared latent ability representation. |
| Outcome: | The proposed framework integrates diagnostic assessment with instructional scaffolding through a shared latent ability representation. |
RaFoLa: A Rationale-Annotated Corpus for Detecting Indicators of Forced Labour (2022.lrec-1)
Copied to clipboard
| Challenge: | Forced labour is the most common type of modern slavery, affecting at least 24.9 million people worldwide. |
| Approach: | They propose to annotate an English corpus for multi-class and multi-label forced labour detection using specialised data from specialised sources. |
| Outcome: | The proposed corpus consists of 989 news articles annotated according to risk indicators defined by the International Labour Organization (ILO). |
Towards Transparent and Explainable Attention Models (2020.acl-main)
Copied to clipboard
Akash Kumar Mohankumar, Preksha Nema, Sharan Narasimhan, Mitesh M. Khapra, Balaji Vasan Srinivasan, Balaraman Ravindran
| Challenge: | Recent studies on interpretability of attention distributions have led to notions of faithful and plausible explanations for a model’s predictions. |
| Approach: | They propose to modify LSTM cells to ensure that the hidden representations learned at different time steps are diverse. |
| Outcome: | The proposed model can provide a faithful explanation if a higher attention weight implies a greater impact on the model’s prediction. |
Leveraging Affirmative Interpretations from Negation Improves Natural Language Understanding (2022.emnlp-main)
Copied to clipboard
| Challenge: | Negation poses a challenge in many natural language understanding tasks . leveraging sentences with negation and affirmative interpretations is beneficial for many tasks involving humans . |
| Approach: | They propose to collect negated sentences and their affirmative interpretations and leverage them to build a plug-and-play neural generator that generates an affirmative interpreter. |
| Outcome: | The proposed method does not require manual effort and does not impact other tasks. |
Does it Make Sense? And Why? A Pilot Study for Sense Making and Explanation (P19-1)
Copied to clipboard
| Challenge: | Existing benchmarks measure common sense knowledge indirectly or without reasoning. |
| Approach: | They propose a benchmark to test whether a system can differentiate natural language statements that make sense from those that do not make sense. |
| Outcome: | The proposed benchmarks show that models trained on large corpora perform better than humans on some benchmarks. |
LIBERTy: A Causal Framework for Benchmarking Concept-Based Explanations of LLMs with Structural Counterfactuals (2026.findings-acl)
Copied to clipboard
| Challenge: | Concept-based explanations quantify how high-level concepts influence model behavior . existing benchmarks rely on costly human-written counterfactuals that serves as imperfect proxy . |
| Approach: | They propose a framework for constructing datasets containing structural counterfactual pairs . they use a structured Causal Model to generate a concept-based explanation . |
| Outcome: | The proposed framework compares concept-based explanations to causal effects estimated from counterfactuals. |
ConvTextTM: An Explainable Convolutional Tsetlin Machine Framework for Text Classification (2022.lrec-1)
Copied to clipboard
| Challenge: | Recent advances in natural language processing (NLP) have reshaped the industry . complexity of such models makes them a “black box” and can cause ethical concerns . |
| Approach: | They propose a convolutional TM architecture that breaks down text into a sequence of fragments . they propose to use a tokenization scheme to bind the tokens to the text fragments. |
| Outcome: | The proposed architecture improves on a set of text fragments and eliminates the need for a corpus-specific vocabulary. |
Interpretable Automatic Fine-grained Inconsistency Detection in Text Summarization (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing factual consistency evaluation approaches for text summarization provide binary predictions and limited insights into the weakness of summarizing systems. |
| Approach: | They propose a task of fine-grained inconsistency detection to detect inconsistent summaries . they represent facts with semantic frames extracted by semantic role labeling and highlight related semantic frames to predict inconsistance . |
| Outcome: | The proposed model outperforms baseline models and provides evidence to support or refute the summary. |
Saliency Learning: Teaching the Model Where to Pay Attention (N19-1)
Copied to clipboard
| Challenge: | Recent work on explanation and interpretation has introduced methods to provide insights toward the model’s behaviour and predictions, but they do not improve the model's reliability. |
| Approach: | They propose to provide explanation training and ensure alignment of model’s explanation with ground truth explanation to ensure the model makes correct predictions for the right reason. |
| Outcome: | The proposed method produces more reliable predictions while delivering better results compared to traditional models. |
Peeking inside the black box: A Commonsense-aware Generative Framework for Explainable Complaint Detection (2023.acl-long)
Copied to clipboard
| Challenge: | Complaining is an expression of negative emotions communicated due to a discrepancy between reality and expectations. |
| Approach: | They propose to use an explainable complaint dataset to generate a commonsense-aware generative framework that can predict the complaint cause, severity level, emotion, and polarity of the text. |
| Outcome: | The proposed model can predict the complaint cause, severity level, emotion, and polarity of the text in addition to detecting whether it is a complaint or not. |
Learning to Faithfully Rationalize by Construction (2020.acl-main)
Copied to clipboard
| Challenge: | Neural models dominate NLP but it remains difficult to know why they make specific predictions for sequential text inputs. |
| Approach: | They propose a model to produce faithful rationales for neural text classification by defining independent snippet extraction and prediction modules. |
| Outcome: | The proposed model produces faithful explanations even when the model is complex and complex. |
Investigating More Explainable and Partition-Free Compositionality Estimation for LLMs: A Rule-Generation Perspective (2026.acl-long)
Copied to clipboard
| Challenge: | Compositional generalization tests focus on output results without considering sample compositionality, resulting in explainability defects. |
| Approach: | They propose a rule-generation perspective for compositionality estimation for LLMs that requires LLM to generate a program as rules for dataset mapping and provides estimates of compositionality using complexity-based theory. |
| Outcome: | The proposed model provides estimates of the compositionality of LLMs using complexity-based theory on a string-to-grid task. |
AbsVis – Benchmarking How Humans and Vision-Language Models “See” Abstract Concepts in Images (2025.emnlp-main)
Copied to clipboard
| Challenge: | Abstract concepts like mercy and peace lack clear visual grounding, and therefore challenge humans and models to provide suitable image representations. |
| Approach: | They propose a dataset of 675 images annotated with 14,175 concept–explanation attributions from humans and two Vision-Language Models where each concept is accompanied by a textual explanation. |
| Outcome: | The proposed dataset compares human and VLM attributions in terms of diversity, abstractness, and alignment, and shows that overlapping concepts are most preferred. |
Rather a Nurse than a Physician - Contrastive Explanations under Investigation (2023.emnlp-main)
Copied to clipboard
| Challenge: | a recent study suggests that contrastive explanations are closer to how humans explain a decision than non-contrastive explanations. |
| Approach: | They analyze four English text-classification datasets to determine whether humans explain in contrast to alternatives. |
| Outcome: | The proposed explanations are closer to how humans explain a decision than non-contrastive explanations. |
Can Explanations Be Useful for Calibrating Black Box Models? (2022.acl-long)
Copied to clipboard
| Challenge: | Existing models are often used as black boxes to adapt to new domains, but there is no single recipe for making them work. |
| Approach: | They propose to use black box models to improve their performance on new domains by leveraging explanations of their behavior. |
| Outcome: | The proposed method improves model generalization performance on two tasks using explanations. |
Investigating the Benefits of Free-Form Rationales (2022.findings-emnlp)
Copied to clipboard
| Challenge: | a recent study shows that crowdsourced rationales provide additional background knowledge to models . a qualitative study shows generated rationale is not as useful for humans as crowdsourced ones . |
| Approach: | They investigate whether crowdsourced rationales provide additional background knowledge to models . they find that ECQA rationale provides additional background information to understand a decision . |
| Outcome: | The results show that ECQA rationales provide additional background knowledge to understand a decision . compared to crowdsourced rationale, generated rationale is not as useful for humans . |
Words as Gatekeepers: Measuring Discipline-specific Terms and Meanings in Scholarly Publications (2023.findings-acl)
Copied to clipboard
| Challenge: | Specialized terminology, or specialized language, can facilitate efficient in-group communication within fields but hinder understanding for out-groups. |
| Approach: | They propose an interpretable approach to measure scholarly jargon from text using word sense induction to identify words that are widespread but overloaded with different meanings across fields. |
| Outcome: | The proposed method is validated by word sense induction and shows that it provides a complementary, yet unique view of jargon alongside word types. |
Enhancing Recommendation Explanations through User-Centric Refinement (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing explanations for user reviews often fail to meet user-centric aspects, reducing their usefulness to users. |
| Approach: | They propose a paradigm that refines initial explanations generated by existing models during the inference stage to enhance their quality in multiple aspects. |
| Outcome: | The proposed model improves explanations generated by existing models during the inference stage to enhance their quality in multiple aspects. |
Explanation Regularisation through the Lens of Attributions (2025.coling-main)
Copied to clipboard
| Challenge: | Explanation regularisation (ER) is a method to guide text classifiers to form their predictions relying on tokens that humans consider plausible. |
| Approach: | They introduce an auxiliary explanation loss to measure how well an input attribution technique's output agrees with human-annotated rationales. |
| Outcome: | The proposed model improves classification performance in out-of-domain (OOD) settings by relying on tokens humans consider plausible. |
Interpretable Semantic Gradients in SSD: A PCA Sweep Approach and a Case Study on AI Discourse (2026.findings-acl)
Copied to clipboard
| Challenge: | Supervised Semantic Differential (SSD) is a mixed quantitative–interpretive method that models how text meaning varies with continuous individual-difference variables . currently no systematic method exists for choosing the number of retained components, introducing avoidable researcher degrees of freedom in the analysis pipeline. |
| Approach: | They propose a PCA sweep procedure that treats dimensionality selection as a joint criterion over representation capacity, gradient interpretability, and stability across nearby values of K. |
| Outcome: | The proposed method is based on a corpus of short posts about artificial intelligence written by Prolific participants who also completed Admiration and Rivalry narcissism scales. |
Competition of Mechanisms: Tracing How Language Models Handle Facts and Counterfactuals (2024.acl-long)
Copied to clipboard
| Challenge: | Existing interpretability research focused on analyzing a single mechanism . et al., 2023) focused on finding how models copy or recall factual knowledge . |
| Approach: | They propose a competition of mechanisms that focuses on the interplay of multiple mechanisms instead of individual mechanisms . they uncover how and where the competition of mechanism happens within LLMs using logit inspection and attention modification methods. |
| Outcome: | The proposed model is based on two interpretability methods, logit inspection and attention modification. |
Bridging the Gap between Expert and Language Models: Concept-guided Chess Commentary Generation and Evaluation (2025.naacl-long)
Copied to clipboard
| Challenge: | Experimental results show that expert models generate accurate, informative and fluent commentary, but are prone to hallucinations due to their limited decision-making capabilities. |
| Approach: | They propose a concept-guided chess commentary generation and a GPT-based Chess Commentary Evaluation to bridge this gap between expert models and large language models. |
| Outcome: | The proposed model is accurate, informative, and fluent. |
Let the CAT out of the bag: Contrastive Attributed explanations for Text (2022.emnlp-main)
Copied to clipboard
| Challenge: | XAI has seen an explosion of interest in explaining black box behavior . contrastive/counterfactual explanations have seen a surge of interest recently . |
| Approach: | They propose a method which provides contrastive explanations for natural language text data with a novel twist by exploiting attribute classifiers. |
| Outcome: | The proposed method outperforms state-of-the-art methods on four benchmark metrics. |
Interpretable Question Answering on Knowledge Bases and Text (P19-1)
Copied to clipboard
| Challenge: | Existing evaluation paradigms for ML based question answering models are lacking . a lack of explanation methods has been proposed for QA models . |
| Approach: | They propose an automatic evaluation paradigm for explanation methods in ML based question answering models . they adapt post hoc explanation methods such as LIME and input perturbation to the model . |
| Outcome: | The proposed evaluation paradigm compares explanation methods with human annotations. |
Explaining Black Box Predictions and Unveiling Data Artifacts through Influence Functions (2020.acl-main)
Copied to clipboard
| Challenge: | Modern deep learning models for NLP are notoriously opaque, and this has motivated efforts to design example-specific approaches to interpret such models. |
| Approach: | They propose to use influence functions to explain models by highlighting important words in input text to provide models with an explanation. |
| Outcome: | The proposed approach is particularly useful for natural language inference, a task in which ‘saliency maps’ may not have clear interpretation. |
Generating Hierarchical Explanations on Text Classification via Feature Interaction Detection (2020.acl-main)
Copied to clipboard
| Challenge: | Existing methods for generating explanations for neural networks ignore feature interactions between words and phrases. |
| Approach: | They propose to build hierarchical explanations by detecting feature interactions by combining words and phrases at different levels of the hierarchy. |
| Outcome: | The proposed method is evaluated on two benchmark datasets, via automatic and human evaluations. |
Understanding How Value Neurons Shape the Generation of Specified Values in LLMs (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Current approaches to interpret value representations are limited by superficial judgments over mechanistic analysis. |
| Approach: | They propose a mechanistic interpretability framework that uses the Schwartz Values Survey to interpret value . they use a dataset that operationalizes four dimensions of universal value through behavioral contexts . |
| Outcome: | The proposed method bridges psychological value frameworks with neuron analysis in large language models. |
Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models (2026.findings-acl)
Copied to clipboard
Hengyuan Zhang, Zhihao Zhang, Ercong Nie, Mingyang Wang, Zunhai Su, Yiwei Wang, Qianli Wang, Shuzhou Yuan, Xufeng Duan, Qibo Xue, Zeping Yu, Chenming Shang, Xiao Liang, Jing Xiong, Hui Shen, Chaofan Tao, Zhengwu Liu, Senjie Jin, Zhiheng Xi, Dongdong Zhang, Sophia Ananiadou, Tao Gui, Ruobing Xie, Hayden Kwok-Hay So, Hinrich Schuetze, Xuanjing Huang, Qi Zhang, Ngai Wong
| Challenge: | Existing literature on mechanistic interpretation (MI) treats it as an observational science, leaving practical applications underexplored. |
| Approach: | They propose a survey structured around the pipeline to identify and improve MI models. |
| Outcome: | The proposed framework enables tangible improvements in Alignment, Capability, and Efficiency. |
Towards Consistent Natural-Language Explanations via Explanation-Consistency Finetuning (2025.coling-main)
Copied to clipboard
| Challenge: | Large language models generate convincing, fluent explanations, but they often generate inconsistent explanations on different inputs. |
| Approach: | They propose a method that adapts large language models to generate more consistent explanations on related examples. |
| Outcome: | The proposed method yields a 10.0% relative explanation consistency improvement across a variety of question-answering datasets and generalizes to 7 out-of-distribution datasets not seen during finetuning (+4.5%) |
Variational Language Concepts for Interpreting Foundation Language Models (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Foundation Language Models (FLMs) have achieved remarkable success in natural language processing. |
| Approach: | They propose a variational Bayesian framework to provide word-level interpretations for FLMs . they propose valc to find optimal language concepts to interpret FLM predictions . |
| Outcome: | Empirical results show that the proposed framework can provide conceptual interpretations for foundation language models. |
Polyjuice: Generating Counterfactuals for Explaining, Evaluating, and Improving Models (2021.acl-long)
Copied to clipboard
| Challenge: | Existing counterfactual generation methods rely on manual labor to create very few counterf actuals or only instantiate limited types of perturbations such as paraphrases or word substitutions. |
| Approach: | They propose a general-purpose counterfactual generator that allows for control over perturbation types and locations. |
| Outcome: | The proposed generator produces diverse sets of realistic counterfactuals that are useful in various applications. |
Towards Faithful Natural Language Explanations: A Study Using Activation Patching in Large Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are capable of generating persuasive Natural Language Explanations (NLEs) however, the faithfulness of these explanations should not be readily trusted at face value. |
| Approach: | They propose to use a causal mediation technique called activation patching to measure the faithfulness of an explanation towards supporting the explained answer. |
| Outcome: | The proposed metric, Causal Faithfulness, quantifies the consistency of causal attributions between explanations and the corresponding model outputs as the indicator of faithfulness. |
So many design choices: Improving and interpreting neural agent communication in signaling games (2023.findings-acl)
Copied to clipboard
| Challenge: | Emergent language games are experimental protocols designed to model how communication may arise among a group of agents. |
| Approach: | They propose to adopt a signaling game in which a sender is exposed to an image and generates a sequence of symbols that is transmitted to a receiver. |
| Outcome: | The proposed language improves when the sender is exposed to an image and generates a sequence of symbols that is transmitted to a receiver. |
Disentangling Transformer Language Models as Superposed Topic Models (2023.emnlp-main)
Copied to clipboard
| Challenge: | Topic Modelling is an established area of research where the quality of a given topic is measured using coherence metrics. |
| Approach: | They propose a weight-based approach to search and disentangle decoder-only TLM by a Wikipedia corpus. |
| Outcome: | The proposed approach can be applied to GPT-2 models and LLaMA models. |
Main Predicate and Their Arguments as Explanation Signals For Intent Classification (2025.naacl-long)
Copied to clipboard
| Challenge: | Intent classification is crucial for conversational agents, and deep learning models perform well in this area due to the lack of suitable benchmark data. |
| Approach: | They propose a technique to augment text samples from intent classification datasets with word-level explanations by marking main predicates and their arguments as explanation signals. |
| Outcome: | The proposed method augments text samples from intent classification datasets with word-level explanations. |
Generating Plausible Counterfactual Explanations for Deep Transformers in Financial Text Classification (2020.coling-main)
Copied to clipboard
| Challenge: | Existing methods for generating textual-based explanations are highly implausible and damage a user’s trust in the automated system. |
| Approach: | They propose a method which first applies robust transformer models on a real-world, up-to-date, self-collected mergers and acquisitions dataset and then generates plausible, post-hoc, counterfactual explanations. |
| Outcome: | The proposed model improves model accuracy and human performance while generating plausible explanations based on human trials. |
PEDANTS: Cheap but Effective and Interpretable Answer Equivalence (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Current short-form QA evaluations lack diverse styles of evaluation data and rely on expensive and slow LLMs. |
| Approach: | They propose a rubric for machine QA that is more stable than an exact match and neural methods. |
| Outcome: | The proposed evaluations improve on the existing short-form QA evaluations using the Trivia community. |
When is Wall a Pared and when a Muro?: Extracting Rules Governing Lexical Selection (2021.emnlp-main)
Copied to clipboard
| Challenge: | Using manual content to learn languages is expensive and time consuming. |
| Approach: | They propose a method for automatically identifying fine-grained lexical distinctions and extracting rules explaining them in a human- and machine-readable format. |
| Outcome: | The proposed method is able to identify fine-grained distinctions and explain them in a human- and machine-readable format. |
Diffusion-CAM: Faithful Visual Explanations for dMLLMs (2026.acl-long)
Copied to clipboard
| Challenge: | Existing Class Activation Mapping methods are ill-suited for interpreting non-autoregressive behaviors of diffusion-based architectures. |
| Approach: | They propose to use a method to generate parallel activation maps by probing intermediate representations in the transformer backbone to capture latent features and their class-specific gradients. |
| Outcome: | Experiments show that Diffusion-CAM significantly outperforms SoTA methods in localization accuracy and visual fidelity. |
A Rose by Any Other Name: LLM-Generated Explanations Are Good Proxies for Human Explanations to Collect Label Distributions on NLI (2025.findings-acl)
Copied to clipboard
| Challenge: | Recent research has shown that explanations provide valuable information for understanding human label variation (HLV) Large language models (LLMs) can approximate HJD from a few human-provided label-explanation pairs, but collecting explanations for every label is still time-consuming. |
| Approach: | They propose to use Large Language Models (LLMs) as annotators to generate model explanations for a few given human labels. |
| Outcome: | The proposed models can generate human-provided explanations from human labels, but they are still time-consuming. |
Detection of Multiple Mental Disorders from Social Media with Two-Stream Psychiatric Experts (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing mental disease detection methods are not backed by domain knowledge and thus fail to produce interpretable results. |
| Approach: | They propose a framework that can learn the shared clues of all diseases while also capturing the specificity of each single disease. |
| Outcome: | Experiments on the detection of 7 diseases show that the proposed model can boost detection performance by more than 10%, especially in relatively rare classes. |
Local Interpretation of Transformer Based on Linear Decomposition (2023.acl-long)
Copied to clipboard
| Challenge: | Existing work on local explanation generation attempts to understand model dynamics on word-level or phraselevel by assigning importance scores on input features. |
| Approach: | They propose to interpret neural networks by linear decomposition by a Transformer model on a single input and a linear decomposing of the output to generate local explanations. |
| Outcome: | The proposed method achieves competitive performance in sentiment classification and machine translation, and fidelity of explanation. |
F1 is Not Enough! Models and Evaluation Towards User-Centered Explainable Question Answering (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing models and evaluation settings have shortcomings regarding the coupling of answer and explanation which might cause serious issues in user experience. |
| Approach: | They propose a hierarchical model and a new regularization term to strengthen the coupling of answer and explanation and two evaluation scores to quantify the couple. |
| Outcome: | The proposed model strengthens the answer-explanation coupling and provides evaluation scores that align with user experience. |
Neuro-Symbolic Sentiment Analysis with Dynamic Word Sense Disambiguation (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Traditional neural network models represent word senses as vectors that are uninterpretable for humans. |
| Approach: | They propose a framework that incorporates word Sense Disambiguation (WSD) by identifying and paraphrasing ambiguous words to improve sentiment predictions. |
| Outcome: | The proposed framework improves sentiment analysis accuracy and interpretability on a downstream task without ground-truth word sense labels. |
Extending Logic Explained Networks to Text Classification (2022.emnlp-main)
Copied to clipboard
| Challenge: | Recent studies have proposed explainable-by-design neural models providing logic explanations for their predictions, but these models favour global explanations, while local ones tend to be noisy and verbose. |
| Approach: | They propose to use LENp to improve local explanations by perturbing input words to improve sensitivity and faithfulness of local explanation. |
| Outcome: | The proposed model provides better local explanations than LIME and is more user-friendly than Lime as attested by a human survey. |
Exploring Interpretability of Independent Components of Word Embeddings with Automated Word Intruder Test (2024.lrec-main)
Copied to clipboard
| Challenge: | Independent Component Analysis (ICA) is an algorithm for finding separate sources in a mixed signal. |
| Approach: | They propose to use ICA to analyze word embeddings to quantify interpretability . they propose to automate word intruder test to quantify the components . |
| Outcome: | The proposed algorithm can be used to find semantic features of words . it can be combined to find words that have features associated with the components . |
Counterfactuals of Counterfactuals: a back-translation-inspired approach to analyse counterfactual editors (2023.findings-acl)
Copied to clipboard
George Filandrianos, Edmund Dervakos, Orfeas Menis Mastromichalakis, Chrysoula Zerva, Giorgos Stamou
| Challenge: | Existing explanations for classifiers are counterfactual or contrastive . lack of universal ground truth for counterf actual edits hinders their evaluation . |
| Approach: | They propose a back translation-inspired evaluation methodology that utilises earlier outputs of the explainer as ground truth proxies to investigate the consistency of explainers. |
| Outcome: | The proposed method can provide valuable insights into the behaviour of predictor and explainer models and infer patterns that would otherwise be obscured. |
LEGO: A Multi-agent Collaborative Framework with Role-playing and Iterative Feedback for Causality Explanation Generation (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Causality explanation generation is a generative task that aims to explain why a given cause-effect pair is true using natural language. |
| Approach: | They propose a multi-agent framework with role-playing and iterative feedback for causality explanation generation. |
| Outcome: | The proposed framework is superior to existing frameworks on WIKIWHY and e-CARE datasets. |
Interpretable Preferences via Multi-Objective Reward Modeling and Mixture-of-Experts (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Reinforcement learning from human feedback (RLHF) is the primary method for aligning large language models with human preferences. |
| Approach: | They propose to train an Absolute-Rating Multi-Objective Reward Model with multi-dimensional absolute-rating data. |
| Outcome: | The proposed model outperforms the LLM-as-a-judge method on RewardBench . it achieves state-of-the-art performance on the benchmark . |
ExecVerify: White-Box RL with Verifiable Stepwise Rewards for Code Execution Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for code execution reasoning are limited by the difficulty of the training data. |
| Approach: | They propose a model that uses reinforcement learning to reward correct answers from execution traces. |
| Outcome: | The proposed model improves pass@1 by up to 5.9% on code generation tasks over strong baselines. |
Enjoy the Salience: Towards Better Transformer-based Faithful Explanations with Word Salience (2021.emnlp-main)
Copied to clipboard
| Challenge: | Pretrained transformer-based language models have demonstrated state-of-the-art predictive performance when adapted into a range of language understanding tasks. |
| Approach: | They propose to use salient information extracted a priori from training data to complement the task-specific information learned by the model during fine-tuning on a downstream task. |
| Outcome: | The proposed model can provide more faithful explanations across four different feature attribution methods compared to vanilla BERT. |
Goal-Driven Explainable Clustering via Language Descriptions (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing formulations neither consider the users’ goals nor explain clusters’ meanings. |
| Approach: | They propose a task formulation that represents both the goal and the explanations as free-form language descriptions. |
| Outcome: | The proposed method produces more accurate and goal-related explanations than previous methods. |
GAProtoNet: A Multi-head Graph Attention-based Prototypical Network for Interpretable Text Classification (2025.coling-main)
Copied to clipboard
| Challenge: | Existing models with black-box nature obscure decision-making process and lack interpretability. |
| Approach: | They propose a multi-head graph attention-based prototypical network that uses a vector and prototypes to learn an interpretable prototypical representation. |
| Outcome: | The proposed model achieves superior results without sacrificing the accuracy of the original black-box LMs. |
Interpretable Word Embeddings via Informative Priors (D19-1)
Copied to clipboard
| Challenge: | Existing word embeddings lack interpretability and are unsupervised . this limitation limits their use within computational social science and digital humanities. |
| Approach: | They propose to use informative priors to create interpretable dimensions for probabilistic word embeddings using a priori model. |
| Outcome: | The proposed models capture latent semantic concepts better than or on-par with the current state of the art while maintaining the simplicity and generalizability of priors. |
Believing without Seeing: Quality Scores for Contextualizing Vision-Language Model Explanations (2026.acl-long)
Copied to clipboard
| Challenge: | Prior work has found that explanations can easily convince users that inaccurate VLM predictions are correct. |
| Approach: | They propose to evaluate two complementary qualities of VLM-generated explanations via two quality scoring functions to improve their accuracy. |
| Outcome: | The proposed explanations improve accuracy on the A-OKVQA, VizWiz, and MMMU-Pro tasks by 11.1%, including a 15.4% reduction in falsely believing incorrect predictions. |
GMEG-EXP: A Dataset of Human- and LLM-Generated Explanations of Grammatical and Fluency Edits (2024.lrec-main)
Copied to clipboard
| Challenge: | Recent work has explored the ability of large language models (LLMs) to generate explanations of existing labeled data. |
| Approach: | They propose a dataset to examine the ability of large language models to explain revisions in sentences by comparing human- and LLM-generated explanations of grammatical and fluency edits to a human evaluation criteria. |
| Outcome: | The proposed explanations address grammatical and fluency edits and are compared with a dataset built from the GMEG (Grammarly Multi-domain Evaluation for GEC) dataset. |
Leveraging LLM-Generated Schema Descriptions for Unanswerable Question Detection in Clinical Data (2025.coling-main)
Copied to clipboard
| Challenge: | Existing methods rely on model uncertainty but lack interpretability and data imbalance. |
| Approach: | They propose a lightweight model that predicts relevant database schemas to detect unanswerable questions, enhancing interpretability and addressing the data imbalance in binary classification tasks. |
| Outcome: | The proposed model improves interpretability and improves accuracy in binary classification tasks. |
ESPRIT: Explaining Solutions to Physical Reasoning Tasks (2020.acl-main)
Copied to clipboard
Nazneen Fatema Rajani, Rui Zhang, Yi Chern Tan, Stephan Zheng, Jeremy Weiss, Aadit Vyas, Abhijit Gupta, Caiming Xiong, Richard Socher, Dragomir Radev
| Challenge: | Neural networks lack the ability to reason about qualitative physics and cannot generalize to scenarios and tasks unseen during training. |
| Approach: | They propose a framework for reasoning about qualitative physics in natural language that generates interpretable descriptions of physical events. |
| Outcome: | The proposed framework generates explanations of how the physical simulation will causally evolve so that an agent or a human can reason about a solution using interpretable descriptions. |
Figure Me Out: A Gold Standard Dataset for Metaphor Interpretation (2020.lrec-1)
Copied to clipboard
| Challenge: | Metaphor comprehension and understanding is a complex cognitive task that requires interpreting metaphors by grasping the interaction between the meaning of their target and source concepts. |
| Approach: | They propose an automatic retrieval approach to annotate verb-noun metaphors in text . they validated their approach by annotating around 1,500 metaphors from tweets . |
| Outcome: | The proposed method reduces the workload on annotators and maintains consistency . it can be used to interpret verb-noun metaphoric expressions in tweets . |
COGS: A Compositional Generalization Challenge Based on Semantic Interpretation (2020.emnlp-main)
Copied to clipboard
| Challenge: | Natural language is characterized by compositionality: meaning of complex expressions is constructed from the meanings of its constituent parts. |
| Approach: | They propose a semantic parsing dataset based on a fragment of English to assess compositional generalization abilities. |
| Outcome: | The proposed model can generalize meanings in a given sentence in 96–99% of the tests, but generalization accuracy is lower and the generalization sensitivity is higher. |
Multi-Domain Explainability of Preferences (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for generating concept-based explanations of preferences are poorly understood. |
| Approach: | They propose a method for generating local and global concept-based explanations of preferences across multiple domains using an LLM. |
| Outcome: | The proposed method outperforms baselines while also being explainable. |
Learning Personalized Alignment for Evaluating Open-ended Text Generation (2024.emnlp-main)
Copied to clipboard
| Challenge: | Traditional evaluation metrics rely heavily on lexical similarity with human-written references, showing poor correlation with human judgments and failing to account for alignment with the diversity of human preferences. |
| Approach: | They propose an interpretable evaluation framework that evaluates alignment with specific human preferences by providing detailed comments and fine-grained scoring. |
| Outcome: | The proposed framework outperforms GPT-4 in Kendall correlation and accuracy with zero-shot reviewers. |
LDIR: Low-Dimensional Dense and Interpretable Text Embeddings with Relative Representations (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing text embeddings with high dimensions are difficult to trace and interpret. |
| Approach: | They propose low-dimensional and interpretable text embeddings with relative representations that encode semantic meanings in a vector space where similar texts are close together in the representation space. |
| Outcome: | The proposed embeddings outperform existing models on multiple tasks with fewer dimensions and are lowdimensional and dense while maintaining interpretability. |
Neighboring Words Affect Human Interpretation of Saliency Explanations (2023.findings-acl)
Copied to clipboard
| Challenge: | Recent studies found that superficial factors such as word length can distort human interpretation of the communicated saliency scores. |
| Approach: | They conduct a user study to examine how the marking of a word’s *neighboring words* affect the explainee’s perception of the word’ s importance in the context of . a saliency explanation. |
| Outcome: | The findings question whether text-based saliency explanations should continue to be communicated at word level and inform future research on alternative methods. |
Label and Explanation Variation in LLM-Based Annotation: a Case Study in Natural Language Inference (2026.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) have shown considerable promise for annotation purposes, but questions remain about their ability to capture human label variation (HLV) label variation is genuine disagreement between annotators observed across NLP tasks. |
| Approach: | They investigate how label and explanation variation manifests within and across LLMs with respect to the Natural Language Inference task. |
| Outcome: | The proposed models generate label distributions similar to humans but exhibit distinct, idiosyncratic judgments and disagreement patterns. |
SConE: Simplified Cone Embeddings with Symbolic Operators for Complex Logical Queries (2023.findings-acl)
Copied to clipboard
| Challenge: | Current geometric-based methods depend on the neural approach to model FOL operators . empirical evidence for explainability is challenging . |
| Approach: | They propose to model conjunction operators using a symbolic modeling approach . they propose to emphasize the essential role of relation projection operator . |
| Outcome: | The proposed method improves answering complex logical queries over previous models. |
NILE : Natural Language Inference with Faithful Natural Language Explanations (2020.acl-main)
Copied to clipboard
| Challenge: | Recent growth in popularity of deep learning models on NLP classification tasks has accompanied the need for generating some form of natural language explanation of predicted labels. |
| Approach: | They propose a novel method which generates labels along with its faithful explanations. |
| Outcome: | The proposed method is more accurate than previously reported methods and has higher sensitivity than previous methods. |
The Language of Trauma: Modeling Traumatic Event Descriptions Across Domains with Explainable AI (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Psychological trauma can manifest following various distressing events, but studies focus on a single aspect of trauma, often neglecting the transferability of findings across different scenarios. |
| Approach: | They propose a language model that fine-tunes a single aspect of trauma to better predict traumatic events across domains. |
| Outcome: | The proposed model outperforms large language models on trauma-related datasets . it also outperformed models on court data, counseling conversations, and forum posts . |
What does a Text Classifier Learn about Morality? An Explainable Method for Cross-Domain Comparison of Moral Rhetoric (2023.acl-long)
Copied to clipboard
Enrico Liscio, Oscar Araque, Lorenzo Gatti, Ionut Constantinescu, Catholijn Jonker, Kyriaki Kalimeri, Pradeep Kumar Murukannaiah
| Challenge: | Existing methods to analyze whether a text classifier learns the domain-specific expression of moral language are lacking. |
| Approach: | They propose a method to compare a supervised classifier’s representation of moral rhetoric across domains by exploring similarities and differences between moral concepts and domains. |
| Outcome: | The proposed method compares a supervised classifier’s representation of moral rhetoric across domains and domains. |
SCDTour: Embedding Axis Ordering and Merging for Interpretable Semantic Change Detection (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods to improve interpretability of SCD often lead to performance degradation . agglomerating axes produces a more refined set of word senses, which improves performance . |
| Approach: | They propose a method that orders and merges interpretable axes to improve SCD performance. |
| Outcome: | The proposed method preserves performance while maintaining high interpretability . it produces a more refined set of word senses, which improves performance . |
Rationales for Sequential Predictions (2021.emnlp-main)
Copied to clipboard
| Challenge: | Sequence models produce accurate predictions, but their decision making processes are hard to explain. |
| Approach: | They propose an efficient algorithm to approximate sequential objective by identifying the most faithful rationales. |
| Outcome: | The proposed algorithm is best at optimizing the sequential objective and provides the most faithful rationales. |
Tracing Logit Trajectories Across Layer Depth: Dataset-Level Explainability for Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Sentence-level explanations miss the bigger picture of how a black-box model behaves across data . a dataset-level analysis that traces the intermediate structure of decision formation is needed . |
| Approach: | They propose a method that aggregates logit updates into a reproducible dataset-level trajectory pattern. |
| Outcome: | The proposed model enables depth-wise explainability across 6 languages and 5 NLP tasks. |
Interpretable Composition Attribution Enhancement for Visio-linguistic Compositional Understanding (2024.emnlp-main)
Copied to clipboard
| Challenge: | Despite promising progress, vision-language models still exhibit significant challenges in understanding visio-linguistic concepts beyond object terms. |
| Approach: | They propose a framework that encourages the model to pay greater attention to composition words denoting relationships and attributes within the text. |
| Outcome: | The proposed framework improves the ability to discern intricate details and construct more sophisticated interpretations of combined visual and linguistic elements. |
Are Human Explanations Always Helpful? Towards Objective Evaluation of Human Natural Language Explanations (2023.acl-long)
Copied to clipboard
| Challenge: | Human-annotated labels and explanations are critical for training explainable NLP models. |
| Approach: | They propose a metric that measures the usefulness of an explanation for model performance at both fine-tuning and inference. |
| Outcome: | The proposed metric can evaluate the quality of human-annotated explanations, while Simulatability falls short. |
TextGenSHAP: Scalable Post-Hoc Explanations in Text Generation with Long Documents (2024.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are difficult to explain and understand due to long input contexts and autoregressive output generation. |
| Approach: | They propose a post-hoc explanation method which incorporates LLM-specific techniques. |
| Outcome: | The proposed method improves retrieval recall and prediction accuracy significantly on open-domain question answering benchmarks. |
MaNtLE: Model-agnostic Natural Language Explainer (2023.emnlp-main)
Copied to clipboard
| Challenge: | Recent research suggests that practitioners prefer examining language explanations that explain sub-groups of examples. |
| Approach: | They propose a model-agnostic natural language explainer that generates faithful explanations of classifier rationale for structured classification tasks. |
| Outcome: | The proposed model-agnostic natural language explainer generates faithful explanations of classifier rationale for structured classification tasks. |
“Seeing the Big through the Small”: Can LLMs Approximate Human Judgment Distributions on NLI from a Few Explanations? (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Human label variation arises when multiple human annotators provide different labels for valid reasons. |
| Approach: | They propose to use crowd workers to represent human judgment distributions or expert linguists to provide detailed explanations for their chosen labels. |
| Outcome: | The proposed model can approximate human judgment distributions using a small number of expert labels and explanations. |
XDetox: Text Detoxification with Token-Level Toxicity Explanations (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for mitigating toxic content are black-box approaches, which results in limitations in modifying toxic tokens. |
| Approach: | They propose a method that integrates token-level toxicity explanations with the masking and infilling detoxification processes. |
| Outcome: | The proposed method outperforms baseline methods in fluency and toxicity reduction. |
PychoAgent: Psychology-driven LLM Agents for Explainable Panic Prediction on Social Media during Sudden Disaster Events (2025.emnlp-main)
Copied to clipboard
Mengzhu Liu, Zhengqiu Zhu, Chuan Ai, Chen Gao, Xinghong Li, Lingnan He, Kaisheng Lai, Yingfeng Chen, Xin Lu, Yong Li, Quanjun Yin
| Challenge: | Social media's rich information content and spatiotemporal granularity provide unique opportunities for emotion prediction and management. |
| Approach: | They propose a Psychology-driven generative Agent framework for explainable panic prediction based on emotion arousal theory. |
| Outcome: | The proposed framework improves panic emotion prediction performance by 13% to 21% compared to baseline models. |
This Reads Like That: Deep Learning for Interpretable Natural Language Processing (2023.emnlp-main)
Copied to clipboard
| Challenge: | In this work, we explore the extension of prototypical networks to natural language processing. |
| Approach: | They propose a weighted similarity measure that enhances the similarity computation by focusing on informative dimensions of pre-trained sentence embeddings. |
| Outcome: | The proposed method improves predictive performance on AG News and RT Polarity datasets and the rationale-based recurrent convolutions. |
Finding Blind Spots in Evaluator LLMs with Interpretable Checklists (2024.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models are increasingly relied upon to evaluate text outputs of other LLMs . however, concerns persist over the accuracy of these assessments and the potential for misleading conclusions. |
| Approach: | They propose a framework to assess the reliability of Large Language Models (LLMs) they propose ' FBI' framework to examine the proficiency of Evaluator LLMs in assessing four critical abilities . |
| Outcome: | The proposed framework assesses the performance of LLMs in text generation tasks. |
AutoPersuade: A Framework for Evaluating and Explaining Persuasive Arguments (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing tools for persuasion are well-equipped to identify which of a pre-existing set of messages is most persuasive, but they do not offer causal evidence on whether or how they have succeeded. |
| Approach: | They propose a framework for identifying topical components of persuasive arguments that are autopersuade. |
| Outcome: | The proposed framework validates the results through human studies and out-of-sample predictions. |
SpecEdit: A Spectral Approach for Multi-Round Knowledge Editing (2026.findings-acl)
Copied to clipboard
| Challenge: | Multi-round knowledge editing suffers from performance degradation as edits accumulate . intrinsic knowledge of model and historical edit memories are naively coupled during editing . SpecEdit improves model editing performance by reducing destructive coupling . |
| Approach: | They propose a spectral-based model editing module that integrates into existing editing methods without altering their original optimization procedures. |
| Outcome: | The proposed model improves performance on multiple LLMs and editing methods. |
Explaining Mixtures of Sources in News Articles (2024.findings-emnlp)
Copied to clipboard
| Challenge: | a recent study shows that language models are essential for long-form article generation. |
| Approach: | They propose a generative process where a source-selection schema is first selected by a journalist, and then sources are chosen based on categories in that schema. |
| Outcome: | The proposed model can predict the most suitable schema given just the headline with reasonable accuracy. |
Logic Rules as Explanations for Legal Case Retrieval (2024.lrec-main)
Copied to clipboard
| Challenge: | Recent efforts to learn explainable legal case retrieval models fail to provide faithful and interpretable explanations for legal cases. |
| Approach: | They propose a framework that uses logic rules to explain legal case retrieval results . they extend benchmarks of LeCaRD and ELAM with manually annotated logic rules . |
| Outcome: | The proposed framework is able to provide faithful explanations for legal case retrieval. |
VISIT: Visualizing and Interpreting the Semantic Information Flow of Transformers (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Recent work in interpretability suggests we can project weights and hidden states of transformer-based language models (LMs) to their vocabulary space, a transformation that makes them more human interpretable. |
| Approach: | They propose a tool to visualize a forward pass of Generative Pre-trained Transformers as an interactive flow graph with nodes representing neurons or hidden states and edges representing interactions between them. |
| Outcome: | The proposed visualization simplifies huge amounts of data into easy-to-read graphs that can reflect the models’ internal processing, uncovering the contribution of each component to the models' final prediction. |
Interpreting Answers to Yes-No Questions in User-Generated Content (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies on yes-no questions outside social media have found that yes and no keywords are rare in answers. |
| Approach: | They propose a corpus of 4,442 yes-no question-answer pairs from twitter . they find that yes and no keywords are rare in answers and poor indicators of correct interpretation . |
| Outcome: | The proposed corpus of 4,442 yes-no question-answer pairs shows that large language models are far from solving the problem. |
Safe-SAIL: Towards a Fine-grained Safety Landscape of Large Language Models via Sparse Autoencoder Interpretation Framework (2026.findings-acl)
Copied to clipboard
Jiaqi Weng, Han Zheng, Hanyu Zhang, Ej Zhou, Qinqin He, Jialing Tao, Hui Xue, Zhixuan Chu, Xiting Wang
| Challenge: | Existing studies on how SAEs derive most fine-grained latent features for safety remain unexplored. |
| Approach: | They propose a framework for interpreting SAE features in safety-critical domains . they train a suite of SAEs with human-readable explanations and systematic evaluations based on pornography, politics, violence, and terror . |
| Outcome: | The proposed framework reduces interpretation cost by 55% and improves safety-critical features. |
Can Multimodal Foundation Models Understand Schematic Diagrams? An Empirical Study on Information-Seeking QA over Scientific Papers (2025.findings-acl)
Copied to clipboard
| Challenge: | MISS-QA is the first benchmark specifically designed to evaluate the ability of models to interpret schematic diagrams within scientific literature. |
| Approach: | They propose an automated evaluation protocol powered by open-source LLMs trained on human-scored data to ensure reliable evaluation. |
| Outcome: | The proposed protocol is powered by open-source LLMs trained on human-scored data. |
Linear Steerability in Language Models: When It Emerges and How It Evolves (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a new framework for steering language models reveals how concepts become linearly separable as training progresses . |
| Approach: | They propose a framework to analyze steerability in language models by using hidden state and representation analysis. |
| Outcome: | The proposed framework reveals how steerability evolves over training . concepts become linearly separable as training progresses, the framework shows . |
CIKT: A Collaborative and Iterative Knowledge Tracing Framework with Large Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Knowledge Tracing (KT) aims to model a student’s learning state over time and predict their future performance. |
| Approach: | They propose a framework that harnesses Large Language Models to enhance both prediction accuracy and explainability by a synergistic optimization loop. |
| Outcome: | The proposed framework improves both prediction accuracy and explainability by using a synergistic optimization loop. |
Rethinking Diverse Human Preference Learning through Principal Component Analysis (2025.findings-acl)
Copied to clipboard
| Challenge: | Decomposed Reward Models extract diverse human preferences from binary comparisons without fine-grained annotations. |
| Approach: | They propose a decomposed reward model that extracts diverse human preferences from binary comparisons without fine-grained annotations. |
| Outcome: | The proposed approach extracts diverse human preferences from binary comparisons without fine-grained annotations. |
Fool Me Once? Contrasting Textual and Visual Explanations in a Clinical Decision-Support Setting (2024.emnlp-main)
Copied to clipboard
Maxime Kayser, Bayar Menzat, Cornelius Emde, Bogdan Bercean, Alex Novak, Abdalá Morgado, Bartlomiej Papiez, Susanne Gaube, Thomas Lukasiewicz, Oana-Maria Camburu
| Challenge: | XAI models are being used in safety-critical domains, but their use is limited due to their limited transparency and insufficient model robustness. |
| Approach: | They evaluated visual, natural language and a combination of both modalities to examine how users use them. |
| Outcome: | The proposed model is more robust and transparent than previous models. |
Evaluating Visual and Cultural Interpretation: The K-Viscuit Benchmark with Human-VLM Collaboration (2025.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to creating inclusive vision-language models rely on human annotators, making it labor-intensive and creating cognitive burdens. |
| Approach: | They propose a semi-automated framework for constructing cultural VLM benchmarks . they use an annotated sample of Korean culture to generate questions . |
| Outcome: | The proposed framework is based on a Korean culture dataset and shows that open-source models lag behind proprietary ones in understanding Korean culture. |
DISCERN: Decoding Systematic Errors in Natural Language for Text Classifiers (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent work proposes automatic methods for identifying and explaining systematic biases using keywords. |
| Approach: | They propose automatic methods for identifying and explaining systematic biases using keywords. |
| Outcome: | The proposed framework improves classifiers by augmenting training sets with synthetically generated instances or annotated examples via active learning. |
Interpretation Meets Safety: A Survey on Interpretation Methods and Tools for Improving LLM Safety (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing surveys focus on interpretation or safety, but safety and understanding are core motivations for interpretation research. |
| Approach: | They propose a framework that connects interpretation methods, enhancements they inform, and tools that operationalize them. |
| Outcome: | The proposed framework summarizes nearly 70 studies at their intersections and concludes with open challenges and future directions. |
CREPE: Rapid Chest X-ray Report Evaluation by Predicting Multi-category Error Counts (2025.emnlp-main)
Copied to clipboard
| Challenge: | CREPE is a metric for rapid, interpretable, and clinically grounded automated chest X-ray report generation. |
| Approach: | They propose to use a domain-specific BERT model fine-tuned with a multi-head regression architecture to predict multi-category error counts across six clinically meaningful categories. |
| Outcome: | CREPE outperforms traditional and recent metrics on a large-scale synthetic dataset of 32,000 annotated report pairs. |
Explicit Bayesian Inference to Uncover the Latent Themes of Large Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) have impressive generative capabilities, yet their inner mechanisms remain largely opaque. |
| Approach: | They propose a variational autoencoder-based neural topic model to interpret LLMs generation process through an explicit Bayesian framework by inferring latent topic variables via variational inference. |
| Outcome: | The proposed model outperforms state-of-the-art topic models on intrinsic measures of coherence and diversity on multiple datasets and shows significant gains on classification and summarization tasks. |
Interpretable Text Embeddings and Text Similarity Explanation: A Survey (2025.emnlp-main)
Copied to clipboard
| Challenge: | Text embeddings are a fundamental component in many NLP tasks, but their interpretation and explanation remain challenging. |
| Approach: | They propose a framework for interpretable text embeddings and text similarity explanation . they characterize the main ideas, approaches, and trade-offs and discuss lessons learned . |
| Outcome: | The proposed methods are compared with existing models and compare them with existing ones. |
Rubrik’s Cube: Testing a New Rubric for Evaluating Explanations on the CUBE dataset (2025.acl-long)
Copied to clipboard
Diana Galvan-Sosa, Gabrielle Gaudeau, Pride Kavumba, Yunmeng Li, Hongyi Gu, Zheng Yuan, Keisuke Sakaguchi, Paula Buttery
| Challenge: | Large-Language Models (LLMs) are increasingly being used in explanation generation tasks due to their unreliability. |
| Approach: | They propose a rubric and a dataset of 26k explanations written and quality-annotated using the rubric by humans and six open- and closed-source LLMs to test their proposed rubric. |
| Outcome: | The proposed rubric and CUBE dataset focuses on reasoning and language tasks and provides the necessary diversity to test it. |
Explain then Rank: Scale Calibration of Neural Rankers Using Natural Language Explanations from LLMs (2025.findings-acl)
Copied to clipboard
| Challenge: | Neural ranking models produce the final document scores, but they are often treated as transient information and only the relative orderings are preserved to produce a ranking. |
| Approach: | They propose to exploit large language models (LLMs) to provide relevance and uncertainty signals for these neural text rankers to produce scale-calibrated scores through Monte Carlo sampling of natural language explanations (NLEs). |
| Outcome: | The proposed approach outperforms previous calibration methods and LLM-based methods for ranking, calibration, and query performance prediction tasks. |
Rapidly Piloting Real-time Linguistic Assistance for Simultaneous Interpreters with Untrained Bilingual Surrogates (2024.lrec-main)
Copied to clipboard
Alvin C. Grissom II, Jo Shoemaker, Benjamin Goldman, Ruikang Shi, Craig Stewart, C. Anton Rytting, Leah Findlater, Jordan Boyd-Graber
| Challenge: | Simultaneous interpretation is a cognitively taxing task, and even seasoned professionals benefit from real-time assistance. |
| Approach: | They propose a simultaneous interpretation task that mimics the cognitive load of interpretation with crowdworker surrogates. |
| Outcome: | The proposed task mimics the cognitive load of interpretation with crowdworker surrogates . the evaluation setup provides consistent results between expert and proxy participants . |
LaTIM: Measuring Latent Token-to-Token Interactions in Mamba Models (2025.acl-long)
Copied to clipboard
| Challenge: | State space models lack interpretability tools for long-context sequence modeling. |
| Approach: | They propose a token-level decomposition method for Mamba-1 and Mamba-2 that enables fine-grained interpretability. |
| Outcome: | The proposed method is able to reveal Mamba’s token-to-token interaction patterns across multiple tasks including translation, copying, and retrieval-based generation. |
Explaining novel senses using definition generation with open language models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | We apply definition generators based on open-weights large language models to create explanations of novel senses. |
| Approach: | They apply open-weights large language models to create explanations of novel senses using target word usages as input. |
| Outcome: | The proposed definition generators perform on par with decoder-only models. |
PrimeX: A Dataset of Worldview, Opinion, and Explanation (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent work shows that an individual's worldview -or beliefs about the overall character of the world -can explain persistent behavioral patterns and correlates with personality, well-being, political, religious, and demographic variables. |
| Approach: | They develop a dataset of public opinion survey data from 858 US residents with written explanations from the respondents for why they hold specific opinions and the Primal World Belief survey for assessing respondent worldview. |
| Outcome: | The proposed model can be used to better represent an individual's belief system and improve opinion prediction. |
CEAES: Bidirectional Reinforcement Learning Optimization for Consistent and Explainable Essay Assessment (2025.acl-long)
Copied to clipboard
| Challenge: | Current automated essay quality assessment systems treat score prediction and feedback generation as separate tasks. |
| Approach: | They propose a bidirectional reinforcement learning framework that jointly optimizes score prediction and feedback generation. |
| Outcome: | The proposed framework outperforms current state-of-the-art models in both scoring and feedback quality. |
From Documents to Segments: A Contextual Reformulation for Topic Assignment (2026.findings-acl)
Copied to clipboard
| Challenge: | Traditional topic modeling treats each document as a single, coherent unit of topic. |
| Approach: | They propose a paradigm that redefines topic assignment at the level of segments . they propose 'segment intrusion task' to extend word intrusion to the span level . |
| Outcome: | The proposed paradigm improves topic purity, interpretability and applicability to multi-theme corpora. |
Semantically Comprehensive Token Pruning in LVLMs via Maximizing Concept Coverage (2026.acl-long)
Copied to clipboard
| Challenge: | Existing visual token pruning methods leverage simple metrics derived from human experience, such as attention or similarity, to rank and select tokens within a highly entangled feature space. |
| Approach: | They propose a novel visual token pruning method that uses a concept-driven paradigm to quantify the Marginal Semantic Gain of each token's contribution to uncovered concepts. |
| Outcome: | The proposed method outperforms state-of-the-art methods in a concept-driven model while maintaining semantic completeness. |
Interpretable Mnemonic Generation for Kanji Learning via Expectation-Maximization (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for mnemonic generation in Japanese are limited in their interpretability due to script differences. |
| Approach: | They propose a method that models the mnemonic construction process as driven by common rules. |
| Outcome: | The proposed method performs well in the cold-start setting for new learners while providing insight into the mechanisms behind effective mnemonic creation. |
An Evaluation Resource for Grounding Translation Errors (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Current fine-grained error analyses do not ground the errors to the reasons why the annotated text spans are erroneous. |
| Approach: | They use a bi-directional grounding scheme to ground erroneous text in two directions . if the error spans of both directions are consistent, the explanation is valid . |
| Outcome: | The proposed grounding process improves translation error detection significantly. |
ELI-Why: Evaluating the Pedagogical Utility of Language Model Explanations (2025.findings-acl)
Copied to clipboard
Brihi Joshi, Keyu He, Sahana Ramnath, Sadra Sabouri, Kaitlyn Zhou, Souti Chattopadhyay, Swabha Swayamdipta, Xiang Ren
| Challenge: | Language models are widely used in education, yet their ability to tailor responses to learners with varied informational needs and knowledge backgrounds remains under-explored. |
| Approach: | They conduct two extensive human studies to assess the utility of language model-generated explanatory answers (explanations) on a benchmark of 13.4K "Why" questions. |
| Outcome: | The proposed model explanations match learners' educational backgrounds only 50% of the time, compared to 79% for lay explanations. |
Colloquial Singaporean English Style Transfer with Fine-Grained Explainable Control (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods for style transfer between Singlish and Standard English lack explainability and fine-grained control. |
| Approach: | They propose a multi-agent framework where large language models act as expert agents for each linguistic aspect. |
| Outcome: | The proposed model enables precise, interpretable transformations, advancing explainability in NLP for Singlish. |
Reheat Nachos for Dinner? Evaluating AI Support for Cross-Cultural Communication of Neologisms (2026.findings-acl)
Copied to clipboard
| Challenge: | Neologisms and emerging slang are central to daily conversation, yet challenging for non-native speakers (NNS) to interpret and use appropriately in cross-cultural communication with native speakers (NS). |
| Approach: | They use AI to learn English neologisms and write messages using the learned word to an NS friend. |
| Outcome: | The proposed model shows that AI Explanation yields the largest gains over no support in NS-rated competence, while contextual appropriateness judgments show indifference across support. |
DVI-DTM: Dual-View Representation Learning for Interpretable Short Text Dynamic Topic Modeling (2026.acl-long)
Copied to clipboard
| Challenge: | Existing dynamic topic modeling methods face semantic ambiguity and interpretation ambiguities when applied to short texts. |
| Approach: | They propose a Dual-View representation learning-based Interpretable short text Dynamic Topic Model to address semantic ambiguity and interpretation ambiguities. |
| Outcome: | The proposed model outperforms the state-of-the-art models in topic alignment and dynamic topic quality metrics while producing highly interpretable topic descriptions. |
Feature-Level Insights into Artificial Text Detection with Sparse Autoencoders (2025.findings-acl)
Copied to clipboard
Kristian Kuznetsov, Laida Kushnareva, Anton Razzhigaev, Polina Druzhinina, Anastasia Voznyuk, Irina Piontkovskaya, Evgeny Burnaev, Serguei Barannikov
| Challenge: | Existing algorithms for AI text detection lack interpretability, limiting their reliability in highstakes applications. |
| Approach: | They extend existing ATD frameworks by using Sparse Autoencoders to extract features from Gemma-2-2b residual stream. |
| Outcome: | The proposed algorithms can extract human-interpretable features from Gemma-2-2b model. |
A Dual-Perspective NLG Meta-Evaluation Framework with Automatic Benchmark and Better Interpretability (2025.acl-long)
Copied to clipboard
| Challenge: | Existing evaluation metrics are insufficient to meet requirements for natural language generation. |
| Approach: | They propose a dual-perspective NLG meta-evaluation framework that focuses on different evaluation capabilities and a method of automatically constructing benchmarks without requiring new human annotations. |
| Outcome: | The proposed framework improves interpretability and provides better performance for 16 representative LLMs. |
DEFT: Demystifying VLN Failures via a Unified Dual-View Explainability Framework for LLM-based Agents (2026.acl-long)
Copied to clipboard
| Challenge: | Existing interpretability methods isolate temporal criticality from feature salience, creating an alignment gap and failing to account for the behavioral instability of black-box agents. |
| Approach: | They propose a unified dual-view framework that jointly analyzes when a decision is pivotal and what visual evidence grounds it. |
| Outcome: | Extensive experiments on MatterPort3D show that DEFT outperforms baselines in both temporal and feature fidelity. |
EXPERT: An Explainable Image Captioning Evaluation Metric with Structured Explanations (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing studies on explainable evaluation metrics generate explanations without standardized criteria and the overall quality of the generated explanations remains unverified. |
| Approach: | They propose a reference-free evaluation metric that provides structured explanations based on fluency, relevance, and descriptiveness. |
| Outcome: | The proposed evaluation template achieves state-of-the-art on benchmark datasets while providing significantly higher-quality explanations than existing metrics. |
The Role of Syntactic Span Preferences in Post-Hoc Explanation Disagreement (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing methods for post-hoc explanations for transformer models disagree with each other . disagreement is often overlooked and the reasons for disagreement are not investigated . |
| Approach: | They propose to use a dynamic *k* approach to estimate syntactic spans to improve agreement between different methods. |
| Outcome: | The proposed method better agrees on syntactic span level, especially for the methods that agree the least with other methods. |
EA-Agent: A Structured Multi-Step Reasoning Agent for Entity Alignment (2026.acl-long)
Copied to clipboard
| Challenge: | Entity alignment (EA) aims to identify entities across different knowledge graphs (KGs) that refer to the same real-world object. |
| Approach: | They propose to use large language models to integrate semantic knowledge into EA to identify entities across different knowledge graphs that refer to the same object. |
| Outcome: | The proposed agent outperforms existing methods and achieves state-of-the-art performance on three benchmark datasets. |
Explaining Differences Between Model Pairs in Natural Language through Sample Learning (2025.emnlp-main)
Copied to clipboard
| Challenge: | a framework that generates faithful natural language explanations of when and how two ML models converge or diverge in their predictions requires access to training data. |
| Approach: | They propose a framework that generates faithful natural language explanations of when and how two ML models converge or diverge in their predictions. |
| Outcome: | The proposed framework generates faithful natural language explanations of when and how two models diverge in their predictions. |
A Causal Lens for Evaluating Faithfulness Metrics (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) offer natural language explanations as an alternative to feature attribution methods for model interpretability, but they may not reflect the model’s truereasoning faithfully. |
| Approach: | They propose a testbed framework for evaluating faithfulness metrics for natural language explanations using diagnosticity and model-editing methods. |
| Outcome: | The proposed framework evaluates faithfulness metrics for natural language explanations on four tasks including fact-checking, analogy, object counting, and multi-hop reasoning. |
LLM-Guided Semantic Bootstrapping for Interpretable Text Classification with Tsetlin Machines (2026.findings-acl)
Copied to clipboard
| Challenge: | Pretrained language models (PLMs) provide strong semantic representations but are costly and opaque. |
| Approach: | They propose a framework that transfers pretrained language models into symbolic form and integrates them into symbolic models. |
| Outcome: | The proposed framework improves interpretability and accuracy across multiple text classification tasks while remaining fully symbolic and efficient. |
XATU: A Fine-grained Instruction-based Benchmark for Explainable Text Updates (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing text editing benchmark datasets contain coarse-grained instructions and lack explainability, resulting in outputs that deviate from intended changes. |
| Approach: | They propose a benchmark specifically designed for fine-grained instruction-based explainable text editing. |
| Outcome: | The proposed benchmark incorporates fine-grained instructions and gold-standard edit explanations. |
Explainable Quantum Program Repair with Verifiable Proof Traces (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to program repair provide only post-hoc, non-verifiable explanations that are not executable or verifiably. |
| Approach: | They propose a framework that couples repair generation with machine-checkable executable explanations for quantum programs where correctness hinges on subtle semantic properties such as circuit equivalence and fidelity preservation. |
| Outcome: | Experiments on QASMBench with mutation-generated quantum program bugs show that the proposed framework improves both semantic precision and explanation faithfulness over baselines that rely on unconstrained or purely natural-language explanations. |
AutoCT: Automating Interpretable Clinical Trial Prediction with LLM Agents (2025.emnlp-main)
Copied to clipboard
| Challenge: | Clinical trials are expensive and time-consuming, and accurate trial prediction is key to advancing medical treatments. |
| Approach: | They propose a framework that combines reasoning capabilities of large language models with the explainability of classical machine learning to generate, evaluate, and refine tabular features without human input. |
| Outcome: | The proposed framework performs better than SOTA methods on clinical trial prediction tasks within a limited number of iterations. |
TravelBehaviorQA: A Benchmark Dataset for Behavioral Interpretation of GPS Trajectories (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in large language models (LLMs) have enabled strong performance on reasoning, summarization, and dialogue tasks across diverse domains. |
| Approach: | They propose a large-scale benchmark dataset that reframes trajectory analysis as a language-based understanding task. |
| Outcome: | The proposed dataset compares GPS trajectories with human-grounded question-answering (QA) pairs. |
Cross-Document Cross-Lingual NLI via RST-Enhanced Graph Fusion and Interpretability Prediction (2025.emnlp-main)
Copied to clipboard
| Challenge: | Despite the development of many subdirections, Cross-Document Cross-Lingual NLI remains largely unexplored. |
| Approach: | They propose a novel paradigm that extends traditional NLI capabilities to multi-document, multilingual scenarios by integrating RST-enhanced graph fusion with interpretability-aware prediction. |
| Outcome: | The proposed method improves on existing models and document-level NLI to multi-document, multilingual scenarios. |
Probabilistic Depression Detection from Textual Time Series (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing models for depression severity estimations lack uncertainty estimates and temporal interpretability. |
| Approach: | They propose a Probabilistic framework for Depression Detection from clinical interview utterance sequences that predicts PHQ-8 scores while modeling calibrated uncertainty. |
| Outcome: | The proposed framework achieves competitive performance among text-only systems and produces well-calibrated intervals. |
Memorization, Emergence, and Explaining Reversal Failures: A Controlled Study of Relational Semantics in LLMs (2026.acl-long)
Copied to clipboard
Yihua Zhu, Qianying Liu, Jiaxin Wang, Fei Cheng, Chaoran Liu, Akiko Aizawa, Sadao Kurohashi, Hidetoshi Shimodaira
| Challenge: | Autoregressive LLMs perform well on relational tasks that require linking entities via relational words, but it is unclear whether they learn the logical semantics of such relations or whether left-to-right order bias is involved. |
| Approach: | They propose a framework that generates text from symmetric/inverse triples and trains autoregressive models from scratch. |
| Outcome: | The proposed framework generates text from symmetric/inverse triples, trains autoregressive models from scratch, and evaluates memorization, logical inference, and in-context generalization to unseen entities. |
NormXLogit: The Head-on-Top Never Lies (2025.emnlp-main)
Copied to clipboard
| Challenge: | NormXLogit is a novel approach for assessing the significance of input tokens based on word embeddings . |
| Approach: | They propose a novel method for assessing the significance of input tokens based on the input and output representations associated with each token. |
| Outcome: | The proposed method outperforms gradient-based methods in faithfulness and offers competitive performance compared to leading architecture-specific techniques. |
A Counterfactual Explanation Framework for Retrieval Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing literature on explainability of information retrieval has focused on illustrating the concept of relevance concerning a retrieval model. |
| Approach: | They propose to add terms to a document to improve its ranking to answer the question of which words played a role in not being favored by a retrieval model. |
| Outcome: | The proposed framework predicts counterfactuals for statistical and deep-learning models. |
Explain the Synth: Interpretable Evaluation of LLM Data Synthesis (2026.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) are increasingly used to generate tabular data. |
| Approach: | They propose a framework that uses a rule-based model as a shared explanatory language to examine the explanation of real versus synthetic data. |
| Outcome: | The proposed framework compares the explanatory structure induced by real versus synthetic data. |
Mechanistic Interpretability of Large-Scale Counting in LLMs through a System-2 Strategy (2026.findings-acl)
Copied to clipboard
Hosein Hasani, Mohammadali Banayeeanzade, Ali Nafisi, Sadegh Mohammadian, Fatemeh Askari, Mobin Bagherian, Amirmohammad Izadi, Mahdieh Soleymani Baghshah
| Challenge: | Large language models exhibit systematic limitations in counting tasks due to depth constraints. |
| Approach: | They propose a method that decomposes large counting tasks into smaller, independent sub-problems that the model can reliably solve. |
| Outcome: | The proposed method surpasses architectural limitations and achieves higher accuracy on large-scale counting tasks. |
Fine-Grained Analysis of Shared Syntactic Mechanisms in Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Recent advances in language models have demonstrated their ability to process linguistic expressions with complex syntactic structures. |
| Approach: | They investigate whether language models employ shared neural mechanisms across different constructions by applying causal interpretability methods at a granular level. |
| Outcome: | The proposed model performance improves on acceptability judgment benchmarks. |