Papers by Yoon Kim
Learning to Decode Collaboratively with Multiple Language Models (2024.acl-long)
Copied to clipboard
| Challenge: | Using a latent variable model, multiple large language models can be trained to collaborate at the token level. |
| Approach: | They propose a method to teach multiple large language models to collaborate by interleaving their generations at the token level. |
| Outcome: | The proposed method improves on instruction-following, domain-specific QA, and reasoning tasks and shows that the model trained with the method exhibits several interesting collaboration patterns. |
Rationale-Guided Retrieval Augmented Generation for Medical Question Answering (2025.naacl-long)
Copied to clipboard
Jiwoong Sohn, Yein Park, Chanwoong Yoon, Sihyeon Park, Hyeon Hwang, Mujeen Sung, Hyunjae Kim, Jaewoo Kang
| Challenge: | Large language models (LLMs) struggle with hallucinations and outdated knowledge. |
| Approach: | They propose a retrieval-augmented generation framework for enhancing the reliability of RAG in biomedical contexts. |
| Outcome: | The proposed framework outperforms the previous best medical RAG model by up to 5.6% across three medical question-answering benchmarks. |
Selective Span-Level Unlearning for Large Language Models (2026.acl-short)
Copied to clipboard
| Challenge: | Existing selective methods that focus on identifying token-level or span-level unlearning targets are misaligning unlearning objectives with the model’s internal behavior. |
| Approach: | They propose a selective method that uses model-intrinsic information to identify token-level or span-level unlearning targets within a text rather than entire sequences. |
| Outcome: | The proposed method achieves comparable unlearning performance while significantly better preserving retained knowledge. |
Deriving Language Models from Masked Language Models (2023.acl-short)
Copied to clipboard
| Challenge: | Masked language models do not define an explicit distribution over language, but they assume that tokens masked out are conditionally independent given the unmasked tokens. |
| Approach: | They propose to use a set of MLM's unary conditionals to construct a fully-connected Markov random field over the input to deduce an explicit joint distribution from MLMs. |
| Outcome: | The proposed method outperforms existing Markov random field-based approaches and outperformed the original model's conditionals. |
Controlling the Focus of Pretrained Language Generation Models (2022.findings-acl)
Copied to clipboard
| Challenge: | Existing mechanisms to control the model's focus are not available for pretrained transformer-based language generation models. |
| Approach: | They propose to augment a pretrained model with trainable "focus vectors" that are directly applied to the model's embeddings while the model itself is kept fixed. |
| Outcome: | The proposed model is able to generate relevant outputs from user-selected highlights while keeping the model fixed. |
Large language models are few-shot clinical information extractors (2022.emnlp-main)
Copied to clipboard
| Challenge: | a long-running goal of clinical NLP is the extraction of important variables trapped in clinical notes. |
| Approach: | They propose to use large language models to tackle diverse clinical extraction tasks . they propose to reannote existing CASI datasets to compare their models with clinical text. |
| Outcome: | The proposed models outperform existing models on few-shot clinical information extraction tasks. |
GrounDial: Human-norm Grounded Safe Dialog Response Generation (2024.findings-eacl)
Copied to clipboard
| Challenge: | Recent conversational AI systems generate unsafe responses agreeing to offensive user input or including toxic content. |
| Approach: | They propose a method where response safety is achieved by grounding responses to commonsense social rules without fine-tuning. |
| Outcome: | The proposed approach is quantitatively and qualitatively safer even without additional data or tuning. |
Automatic Creation of Named Entity Recognition Datasets by Querying Phrase Representations (2023.acl-long)
Copied to clipboard
| Challenge: | Named entity recognition models rely on domain-specific dictionaries provided by experts . however, such dictionary sets are infeasible in many domains where they do not exist . |
| Approach: | They propose a framework that generates NER datasets with high-coverage pseudo-dictionaries . phrase retrieval models are used to retrieve popular entities rather than rare ones . |
| Outcome: | The proposed framework outperforms the previous best model by an average F1 score of 4.7 across five NER benchmark datasets. |
ETHIC: Evaluating Large Language Models on Long-Context Tasks with High Information Coverage (2025.naacl-long)
Copied to clipboard
| Challenge: | Existing evaluation methods do not assess whether large language models fully utilize contextual information. |
| Approach: | They introduce a new metric to assess LLMs' ability to fully utilize contextual information. |
| Outcome: | The proposed benchmark comprises 1,986 test instances spanning four long-context tasks with high IC scores in the domains of books, debates, medicine, and law. |
Propagate-Selector: Detecting Supporting Sentences for Question Answering via Graph Neural Networks (2020.lrec-1)
Copied to clipboard
| Challenge: | Existing question-answering models do not require reasoning across sentences in the given context (passage). |
| Approach: | They propose a graph neural network that propagates information over sentences to understand information that cannot be inferred when considering sentences in isolation. |
| Outcome: | The proposed approach obtains the best performance compared to the widely used answer-selection models that do not consider the intersentential relationship. |
LangSAE Editing: Improving Multilingual Information Retrieval via Post-hoc Language Identity Removal (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for dense retrieval in multilingual environments encode language identity alongside semantics. |
| Approach: | They propose a method that trains on pooled embeddings to remove language-identity signal directly in vector space. |
| Outcome: | The proposed method improves ranking quality and cross-language coverage across multiple languages with especially strong gains for script-distinct languages. |
Beyond Task-Oriented and Chitchat Dialogues: Proactive and Transition-Aware Conversational Agents (2025.emnlp-main)
Copied to clipboard
| Challenge: | Current efforts to bridge the two modes of interaction are reactive, focusing on responding to user inputs rather than coordinating dialogue flows. |
| Approach: | They propose a dataset designed for transition-aware dialogue modeling that incorporates structurally diverse and integrated mode flows. |
| Outcome: | The proposed dataset outperforms baseline models in intent detection and mode transition handling. |
Enhancing Time Awareness in Generative Recommendation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing models focus on sequential order of items and neglect to handle temporal dynamics . existing models neglect to capture hidden user preferences via various temporal signals . |
| Approach: | They propose a model that generates recommendations into a text-to-text generation task . they introduce Time-aware Prompting and Trend-awful Inference . |
| Outcome: | The proposed model outperforms state-of-the-art models with gains of 15.4% and 14.3% . it is based on time-aware Prompting and Trend-awful Inference . |
A Survey on LLM-based Conversational User Simulation (2026.eacl-long)
Copied to clipboard
Bo Ni, Yu Wang, Leyao Wang, Branislav Kveton, Franck Dernoncourt, Yu Xia, Hongjie Chen, Reuben Luera, Samyadeep Basu, Subhojyoti Mukherjee, Puneet Mathur, Nesreen K. Ahmed, Junda Wu, Li Li, Huixin Zhang, Ruiyi Zhang, Tong Yu, Sungchul Kim, Jiuxiang Gu, Zhengzhong Tu, Alexa Siu, Zichao Wang, Seunghyun Yoon, Nedim Lipka, Namyong Park, Zihao Lin, Trung Bui, Yue Zhao, Tyler Derr, Ryan A. Rossi
| Challenge: | Recent advances in large language models (LLMs) have enabled high-fidelity generation of synthetic user conversation. |
| Approach: | They propose a taxonomy covering user granularity and simulation objectives . they analyze core techniques and evaluation methodologies to help them understand the latest developments . |
| Outcome: | The proposed model enables high-fidelity generation of synthetic user conversation. |
SelFusion: Self-distillation for Diffusion Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Existing knowledge distillation methods for autoregressive large language models (LLMs) are not effective for reducing generation quality, but they can be useful for real-time applications. |
| Approach: | They propose a self-distillation framework that allows for effective KD without external teacher . they propose to use two modes of knowledge distillation to determine distillation direction . |
| Outcome: | The proposed framework outperforms existing methods with external teachers on instruction-following tasks. |
Compound Probabilistic Context-Free Grammars for Grammar Induction (P19-1)
Copied to clipboard
| Challenge: | Existing approaches to grammar induction have resorted to manually-engineered features and auxiliary objectives to induce the desired structures. |
| Approach: | They propose a formalization of the grammar induction problem that models sentences as being generated by a compound probabilistic context free grammar. |
| Outcome: | Experiments on English and Chinese show that the proposed approach is more efficient than other methods. |
Sequence-Level Mixed Sample Data Augmentation (2020.emnlp-main)
Copied to clipboard
| Challenge: | Despite their empirical success, neural networks still have difficulty capturing compositional aspects of natural language. |
| Approach: | They propose a data augmentation approach to encourage compositional behavior in neural networks . they propose to softly combine input/output sequences from the training set . |
| Outcome: | The proposed approach yields 1.0 BLEU improvement on translation datasets over baselines. |
Exploring the Practicality of Generative Retrieval on Dynamic Corpora (2024.emnlp-main)
Copied to clipboard
| Challenge: | a lack of comprehensive comparison between GR and Dual Encoders in IR systems is limiting . GR is more adaptable to evolving knowledge (4–11%), robust in learning knowledge with temporal information, and efficient in terms of inference FLOPs (x2), indexing time (x6) and storage footprint (x4) |
| Approach: | They propose to use autoregressive language models to perform information retrieval (IR) their results highlight the potential of GR for future use in practical IR systems . |
| Outcome: | The proposed model is more adaptable to evolving knowledge (4–11%), robust in learning knowledge with temporal information, efficient inference FLOPs (x2), indexing time (x6), and storage footprint (x4) compared to the most common model, Dual Encoder (DE). |
Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks (2024.naacl-long)
Copied to clipboard
Zhaofeng Wu, Linlu Qiu, Alexis Ross, Ekin Akyürek, Boyuan Chen, Bailin Wang, Najoung Kim, Jacob Andreas, Yoon Kim
| Challenge: | Recent language models possess impressive performance across a wide range of tasks . however, they often rely on narrow, non-transferable procedures for task-solving . |
| Approach: | They propose to evaluate language models using "counterfactual" task variants that deviate from standard tasks. |
| Outcome: | The proposed framework shows that language models perform better on a wide range of tasks compared to the default conditions. |
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. |
PR-MCS: Perturbation Robust Metric for MultiLingual Image Captioning (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing image captioning metrics are vulnerable to lexical perturbations, but they are not robust to such perturbations. |
| Approach: | They propose a perturbation-robust multilingual CLIPScore which is a reference-free image captioning metric for multiple languages. |
| Outcome: | The proposed metric outperforms baseline metrics in capturing lexical noise of all various perturbation types in all five languages while maintaining a strong correlation with human judgments. |
Rare Tokens Degenerate All Tokens: Improving Neural Text Generation via Adaptive Gradient Gating for Rare Token Embeddings (2022.acl-long)
Copied to clipboard
| Challenge: | Recent studies have determined that the learned token embeddings of large-scale neural language models are degenerated to be anisotropic with a narrow-cone shape. |
| Approach: | They propose a method to degenerate the learning gradient for rare token embeddings by gating the specific part of the gradient for all tokens during training stage. |
| Outcome: | The proposed method improves the performance of the models but lacks the training dynamics needed to solve the representation degeneration problem. |
Parallel-SFT: Improving Zero-Shot Cross-Programming-Language Transfer for Code RL (2026.findings-acl)
Copied to clipboard
Zhaofeng Wu, Shiqi Wang, Boya Peng, Anuj Kumar Goyal, Melanie Kambadur, Sebastian Ruder, Yoon Kim, Chloe Bi
| Challenge: | Modern language models demonstrate impressive coding capabilities in common programming languages (PLs) but their performance in lower-resource PLs is often limited by training data availability. |
| Approach: | They propose a zero-shot cross-programming-language transfer task for code RL . they propose RL training in a source PL fails to improve performance on other target PLs . |
| Outcome: | The proposed approach improves transferability in Llama-3.1 code generation on parallel-stack model . it also improves performance on other target PLs, compared to single-PL SFT . |
Entailment as Robust Self-Learner (2023.acl-long)
Copied to clipboard
| Challenge: | Recent studies have found that entailment pretraining benefits weakly supervised fine-tuning. |
| Approach: | They propose a prompting strategy that formulates different NLU tasks as contextual entailment and propose an algorithm for better pseudo-labeling quality in self-training. |
| Outcome: | The proposed approach improves the zero-shot adaptation performance on downstream tasks. |
Selective Test-Time Debiasing for CLIP via Reward Gating (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for debiasing use uniform bias corrections across all input queries . weak debiases retains bias in sensitive queries, while weak dealiases in biased ones . |
| Approach: | They propose a framework that selectively applies debiasing based on input sensitivity . RG-TTA adaptively triggers fairness regularization based upon bias sensitivity of each input . |
| Outcome: | Experiments show that debiasing improves zero-shot performance while maintaining fairness . weak debiased queries distort semantically meaningful information while weak ones fail to mitigate stereotypes . |
FaithUn: Toward Faithful Forgetting in Language Models by Investigating the Interconnectedness of Knowledge (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for unlearning undesirable knowledge have overlooked complexity and interconnectedness of knowledge, authors say . previous studies have neglected the complex nature of knowledge and neglected its internal dependencies. |
| Approach: | They propose a new concept called superficial unlearning to evaluate faithfulness of unlearning in knowledge QA settings. |
| Outcome: | The proposed method shows significant effectiveness in real-world knowledge QA settings. |
KPQA: A Metric for Generative Question Answering Using Keyphrase Weights (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing n-gram similarity metrics fail to discriminate the incorrect answers due to the free-form of the answer. |
| Approach: | They propose a new metric that assigns different weights to each token via keyphrase prediction to judge the correctness of GenQA. |
| Outcome: | The proposed metric has a significantly higher correlation with human judgments than existing metrics in various datasets. |
Search Augmented Instruction Learning (2023.findings-emnlp)
Copied to clipboard
Hongyin Luo, Tianhua Zhang, Yung-Sung Chuang, Yuan Gong, Yoon Kim, Xixin Wu, Helen Meng, James Glass
| Challenge: | Large language models (LLMs) have been significantly improved by instruction fine-tuning, but still lack transparency and the ability to utilize up-to-date knowledge and information. |
| Approach: | They propose a search-augmented instruction learning model which grounds the language generation and instruction following abilities on complex search results generated by in-house and external search engines. |
| Outcome: | The proposed model outperforms plain LLMs on zero-shot language tasks and can generate both natural and programming languages following natural language guidance and requests. |
From Selection to Generation: A Survey of LLM-based Active Learning (2025.acl-long)
Copied to clipboard
Yu Xia, Subhojyoti Mukherjee, Zhouhang Xie, Junda Wu, Xintong Li, Ryan Aponte, Hanjia Lyu, Joe Barrow, Hongjie Chen, Franck Dernoncourt, Branislav Kveton, Tong Yu, Ruiyi Zhang, Jiuxiang Gu, Nesreen K. Ahmed, Yu Wang, Xiang Chen, Hanieh Deilamsalehy, Sungchul Kim, Zhengmian Hu, Yue Zhao, Nedim Lipka, Seunghyun Yoon, Ting-Hao Kenneth Huang, Zichao Wang, Puneet Mathur, Soumyabrata Pal, Koyel Mukherjee, Zhehao Zhang, Namyong Park, Thien Huu Nguyen, Jiebo Luo, Ryan A. Rossi, Julian McAuley
| Challenge: | Large Language Models (LLMs) have been used for selection and training of data for active learning. |
| Approach: | They propose an intuitive taxonomy that categorizes LLM-based active learning techniques and discuss the transformative roles they can play in the active learning loop. |
| Outcome: | The proposed model can generate entirely new data instances and provide more cost-effective annotations with fewer labeled data instances. |
Can Large Language Models Keep Up? Benchmarking Online Adaptation to Continual Knowledge Streams (2026.acl-long)
Copied to clipboard
Jiyeon Kim, Hyunji Lee, Dylan Zhou, Sue Hyun Park, Seunghyun Yoon, Trung Bui, Franck Dernoncourt, Sungmin Cha, Minjoon Seo
| Challenge: | Existing models and agentic memory systems fail to adapt robustly to OAKS, demonstrating delays in state-tracking and susceptibility to distraction within streaming environments. |
| Approach: | They propose a benchmark to evaluate models' ability to adapt to changing knowledge over streaming . they use two datasets to analyze how facts evolve over time . |
| Outcome: | The proposed benchmark evaluates models in an online adaptation setting over streaming, continually updating knowledge. |
Is Safety Standard Same for Everyone? User-Specific Safety Evaluation of Large Language Models (2025.findings-emnlp)
Copied to clipboard
Yeonjun In, Wonjoong Kim, Kanghoon Yoon, Sungchul Kim, Mehrab Tanjim, Sangwu Park, Kibum Kim, Chanyoung Park
| Challenge: | Extensive benchmarks evaluate LLM safety relying heavily on general standards . no benchmark datasets exist to evaluate the user-specific safety of LLMs . |
| Approach: | a new benchmark is designed to assess user-specific aspect of LLM safety . authors propose a simple remedy based on chain-of-thought to improve user-specified safety. |
| Outcome: | a new benchmark assesses the user-specific aspect of LLM safety . the proposed solution improves user-specified safety by chain-of-thought . |
Simple Hardware-Efficient PCFGs with Independent Left and Right Productions (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to scaling dense PCFGs to thousands of nonterminals have shown to be beneficial for unsupervised parsing, but they still perform poorly as a language model and as an unsupervised model. |
| Approach: | They propose a simple PCFG formalism with independent left and right productions that scales more effectively as a language model and as an unsupervised parser. |
| Outcome: | The proposed formalism scales better as a language model and as an unsupervised parser despite imposing a stronger independence assumption compared to low-rank approaches. |
Exploring the Potential of LLMs as Personalized Assistants: Dataset, Evaluation, and Analysis (2025.acl-long)
Copied to clipboard
| Challenge: | Personalized AI assistants are a challenging application that intertwines multiple problems in LLM research. |
| Approach: | They propose a Llama-3.2-based automated evaluation model that matches human preferences to a conversational dataset. |
| Outcome: | HiCUPID provides a conversational dataset tailored for personalization . the evaluation model closely mirrors human preferences, the researchers show . |
DPLoRA: A Dual-Pruning Framework based on ILP Optimization and Progressive Pruning for Parameter-Efficient LoRA Fine-Tuning (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) require computational resources for fine-tuning. |
| Approach: | They propose a framework that optimizes rank allocation via two stages . they propose an initial pruning stage and a progressive pruning stage . |
| Outcome: | The proposed framework outperforms existing PEFT baselines on GLUE and instruction-following tasks while reducing training time and trainable parameters by over 80%. |
Debiasing Online Preference Learning via Preference Feature Preservation (2025.findings-acl)
Copied to clipboard
| Challenge: | Recent preference learning frameworks for large language models (LLMs) simplify human preferences with binary pairwise comparisons and scalar rewards. |
| Approach: | They propose a preference feature preservation framework that preserves the distribution of human preference features and maps them throughout the online preference learning process. |
| Outcome: | The proposed framework maintains the distribution of human preference features and utilizing such rich signals throughout the online preference learning process. |
Parameter-Efficient Transfer Learning with Diff Pruning (2021.acl-long)
Copied to clipboard
| Challenge: | Pretrained networks are difficult to deploy for multiple tasks in storage-constrained settings. |
| Approach: | Diff pruning enables parameter-efficient transfer learning that scales well with new tasks. |
| Outcome: | Diff pruning can match the performance of finetuned baselines on the GLUE benchmark while only modifying 0.5% of the pretrained model’s parameters per task. |
Being Kind Isn’t Always Being Safe: Diagnosing Affective Hallucination in LLMs (2026.findings-eacl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are increasingly engaged in emotionally vulnerable conversations that extend beyond information seeking to moments of personal distress. |
| Approach: | They propose AHaBench, a benchmark of 500 mental-health-related prompts with expert-informed reference responses, evaluated along three dimensions: Emotional Enmeshment, Illusion of Presence, and Fostering Overdependence. |
| Outcome: | The proposed model is based on 500 mental-health-related prompts with expert-informed reference responses and a 5K-instance preference dataset enabling direct preference optimization (DPO) for alignment with emotionally responsible behavior. |
Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignment (2024.emnlp-main)
Copied to clipboard
| Challenge: | Multilingual human preference data are difficult to obtain at scale, making it challenging to extend this framework to diverse languages. |
| Approach: | They propose a method where a reward model is trained on preference data in one source language and applied to other target languages. |
| Outcome: | The proposed approach is effective under comprehensive evaluation settings, including human evaluation. |
TLCR: Token-Level Continuous Reward for Fine-grained Reinforcement Learning from Human Feedback (2024.findings-acl)
Copied to clipboard
Eunseop Yoon, Hee Suk Yoon, SooHwan Eom, Gunsoo Han, Daniel Nam, Daejin Jo, Kyoung-Woon On, Mark Hasegawa-Johnson, Sungwoong Kim, Chang Yoo
| Challenge: | Existing approaches to provide token-level rewards fail to account for varying degrees of preference inherent to each token. |
| Approach: | They propose a reward model that uses a discriminator to assign token-based continuous rewards to each token considering the context. |
| Outcome: | Extensive experiments show that the proposed reward model improves on open-ended language generation benchmarks. |
MDCR: A Dataset for Multi-Document Conditional Reasoning (2024.findings-emnlp)
Copied to clipboard
| Challenge: | ConditionalQA is limited to questions on single documents, neglecting harder cases that may require *cross-document reasoning* and *optimization*. |
| Approach: | They propose to use a dataset to evaluate models' ability to answer eligibility questions on single documents. |
| Outcome: | The proposed dataset can reflect real-world challenges and serve as a test bed for complex conditional reasoning that requires optimization. |
Do MLLMs Capture How Interfaces Guide User Behavior? A Benchmark for Multimodal UI/UX Design Understanding (2026.acl-long)
Copied to clipboard
Jaehyun Jeon, Min Soo Kim, Janghan Yoon, Sumin Shim, Yejin Choi, Hanbin Kim, Dae Hyun Kim, Youngjae Yu
| Challenge: | Recent studies focus on surface-level features, overlooking how design choices influence user behavior at scale. |
| Approach: | They propose a benchmark for multimodal understanding of how UI/UX design affects user behavior built on 300 real-world UI image pairs from industry A/B tests. |
| Outcome: | The proposed benchmarks show that models exhibit limited understanding of the behavioral impact of UI/UX design. |
Fast Matrix Multiplications for Lookup Table-Quantized LLMs (2024.findings-emnlp)
Copied to clipboard
| Challenge: | generative (tokenby-token) inference is memory-bound and requires a large amount of memory to perform. |
| Approach: | They propose a lookup table engine for weight-quantized large language models that uses offline restructuring of the quantized weight matrix to minimize bit manipulations associated with unpacking. |
| Outcome: | The proposed kernel can be 2-4x faster than existing GEMM kernels while achieving performance gains of 1.5 to 2 times. |
VLind-Bench: Measuring Language Priors in Large Vision-Language Models (2025.findings-naacl)
Copied to clipboard
| Challenge: | Large Vision-Language Models suffer from a problem known as language prior . such language priors can lead to undesirable biases and hallucinations when dealing with images that are out of distribution. |
| Approach: | They propose a benchmark to measure the language priors of Large Vision-Language Models. |
| Outcome: | The proposed benchmark is the first specifically designed to measure the language priors, or blindness, of LVLMs. |
ToDi: Token-wise Distillation via Fine-Grained Divergence Control (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) offer impressive performance but are impractical for resource-constrained deployment due to high latency and energy consumption. |
| Approach: | They propose a method that adaptively combines FKL and RKL per token using a sigmoid-based weighting function derived from the teacher-student probability log-ratio. |
| Outcome: | The proposed method outperforms baselines using uniform or less granular strategies across instruction-following benchmarks. |
EdiText: Controllable Coarse-to-Fine Text Editing with Diffusion Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods for text editing have been proposed for various types of data with diverse attributes. |
| Approach: | They propose a novel text editing method that modifies reference text to desired attributes at various scales. |
| Outcome: | The proposed method is capable of making precise adjustments within the desired range while maintaining the accuracy of the reference text. |
Generating Diverse Hypotheses for Inductive Reasoning (2025.naacl-long)
Copied to clipboard
| Challenge: | Recent studies suggest that large language models (LLMs) can engage in inductive reasoning by sampling multiple hypotheses about the rules and selecting the one that best explains the observations. |
| Approach: | They propose to increase the temperature parameter to enhance diversity by sampling multiple hypotheses and selecting the one that best explains the observations. |
| Outcome: | The proposed method improves diversity while maintaining text quality while increasing temperature. |
Semantic Token Reweighting for Interpretable and Controllable Text Embeddings in CLIP (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Despite the varying significance of textual elements within a sentence depending on the context, efforts to account for variation of importance in constructing text embeddings have been lacking. |
| Approach: | They propose a framework for Semantic Token Reweighting to build Interpretable text embeddings which incorporates controllability as well. |
| Outcome: | The proposed framework improves the text encoding process in CLIP by differentially weighting semantic elements based on contextual importance, enabling finer control over emphasis responsive to data-driven insights and user preferences. |
Natural Language Embedded Programs for Hybrid Language Symbolic Reasoning (2024.findings-naacl)
Copied to clipboard
Tianhua Zhang, Jiaxin Ge, Hongyin Luo, Yung-Sung Chuang, Mingye Gao, Yuan Gong, Yoon Kim, Xixin Wu, Helen Meng, James Glass
| Challenge: | Existing methods for surfacing symbolic reasoning capabilities are limited to narrow tasks . arithmetic computations are unnatural to perform in pure language space, and hence present difficulties for LLMs. |
| Approach: | They propose a natural language embedded program framework for solving symbolic reasoning tasks. |
| Outcome: | The proposed framework improves on strong baselines across math and symbolic reasoning, text classification, question answering, and instruction following tasks. |
DiZiNER: Disagreement-guided Instruction Refinement via Simulating Pilot Annotation for Zero-shot Named Entity Recognition (2026.acl-long)
Copied to clipboard
| Challenge: | Large language models have advanced information extraction (IE) by enabling zero-shot and few-shot named entity recognition (NER) but their outputs still show persistent and systematic errors. |
| Approach: | They propose a framework that simulates the pilot annotation process and employs LLMs as both annotators and supervisors to refine model disagreements. |
| Outcome: | Using a pilot annotation process, the proposed framework outperforms its supervisor model on 18 benchmarks. |
On the Same Wavelength? Evaluating Pragmatic Reasoning in Language Models across Broad Concepts (2025.emnlp-main)
Copied to clipboard
| Challenge: | Language models (LMs) are increasingly used as conversational agents because of their pragmatic reasoning abilities. |
| Approach: | They propose an evaluation framework derived from *Wavelength*, a popular communication game where a speaker and a listener communicate about a broad range of concepts in a granular manner. |
| Outcome: | The proposed evaluation framework outperforms direct and Chain-of-Thought (CoT) prompting on language comprehension and language production tasks. |
DiffCSE: Difference-based Contrastive Learning for Sentence Embeddings (2022.naacl-main)
Copied to clipboard
Yung-Sung Chuang, Rumen Dangovski, Hongyin Luo, Yang Zhang, Shiyu Chang, Marin Soljacic, Shang-Wen Li, Scott Yih, Yoon Kim, James Glass
| Challenge: | Recent work shows that finetuning pretrained models with contrastive learning makes it possible to learn good sentence embeddings without labeled data. |
| Approach: | They propose an unsupervised contrastive learning framework for learning sentence embeddings . they use a masked language model to mask out the edited sentence . |
| Outcome: | The proposed framework outperforms SimCSE on semantic textual similarity tasks by 2.3 absolute points. |
QuDAR: Query-Wise Dual-Perspective Adaptive Retrieval (2026.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to grounding large language models rely on static weights and a static retrieval component. |
| Approach: | They propose a dual-perspective adaptive retrieval framework that adapts along two perspectives: retriever type (sparse vs. dense) and query format (original v. expanded). |
| Outcome: | The proposed framework adapts along two perspectives: retriever type (sparse vs. dense) and query format (original v. expanded). |
Unleashing Multi-Hop Reasoning Potential in Large Language Models through Repetition of Misordered Context (2025.findings-naacl)
Copied to clipboard
| Challenge: | Multi-hop reasoning requires multi-step reasoning based on supporting documents within a given context. |
| Approach: | They propose a method that prompts the model by repeatedly presenting the context. |
| Outcome: | The proposed method improves the F1 score by 30%p on multi-hop QA tasks and increases accuracy by 70%p on a synthetic task. |
Does Your Voice Assistant Remember? Analyzing Conversational Context Recall and Utilization in Voice Interaction Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in multi-turn voice interaction models have improved user-model communication, but whether open-source models share this ability remains unexplored. |
| Approach: | They propose to use ContextDialog to evaluate open-source interaction models' ability to recall past utterances to identify key limitations. |
| Outcome: | The proposed model retains and recalls past utterances better than closed-source models, but still struggles with questions about past . findings highlight key limitations in open-source model and suggest ways to improve memory retention and retrieval robustness. |
Learning to See through Sound: From VggCaps to Multi2Cap for Richer Automated Audio Captioning (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing AAC datasets suffer from short and simplistic captions, limiting expressiveness and semantic depth. |
| Approach: | They propose a multi-modal dataset that pairs audio with corresponding video and leverages large language models to generate rich, descriptive captions. |
| Outcome: | The proposed framework outperforms existing benchmarks in caption length, lexical diversity, and human-rated quality. |
Simple Questions Generate Named Entity Recognition Datasets (2022.emnlp-main)
Copied to clipboard
| Challenge: | Recent named entity recognition models rely on human-annotated datasets . however, in-domain dictionaries and sentences are often unavailable or expensive to construct for many entity types. |
| Approach: | They propose an ask-to-generate approach which automatically generates NER datasets by asking natural language questions to an open-domain question answering system. |
| Outcome: | The proposed model outperforms the previous best model by 19.5 F1 score on six benchmarks and achieves state-of-the-art performance. |
Assessing LLM Reasoning Steps via Principal Knowledge Grounding (2025.findings-emnlp)
Copied to clipboard
Hyeon Hwang, Yewon Cho, Chanwoong Yoon, Yein Park, Minju Song, Kyungjae Lee, Gangwoo Kim, Jaewoo Kang
| Challenge: | Step-by-step reasoning has become a standard approach for large language models to tackle complex tasks. |
| Approach: | They propose a framework that assesses the knowledge grounding of intermediate reasoning by using a large-scale repository of atomic knowledge essential for reasoning. |
| Outcome: | The evaluation suite identifies missing or misapplied knowledge elements and provides crucial insights for uncovering fundamental reasoning deficiencies in LLMs. |
Difficulty-Focused Contrastive Learning for Knowledge Tracing with a Large Language Model-Based Difficulty Prediction (2024.lrec-main)
Copied to clipboard
Unggi Lee, Sungjun Yoon, Joon Seo Yun, Kyoungsoo Park, YoungHoon Jung, Damji Stratton, Hyeoncheol Kim
| Challenge: | Existing studies have focused on incorporating the difficulty information into knowledge tracing models, but few studies have explored the potential of difficulty estimation. |
| Approach: | They propose a difficulty-centered contrastive learning method and a Large Language Model-based framework for difficulty prediction to improve the performance of knowledge tracing models. |
| Outcome: | The proposed methods demonstrate enhanced performance of knowledge tracing models while ignoring the complex relationship between language and difficulty. |
LangNav: Language as a Perceptual Representation for Navigation (2024.findings-naacl)
Copied to clipboard
| Challenge: | Existing approaches to vision-and-language navigation use visual features as the perceptual representation of a visual representation of an agent's egocentric panoramic view. |
| Approach: | They propose to use off-the-shelf vision systems to convert an agent’s egocentric panoramic view into natural language descriptions. |
| Outcome: | The proposed approach improves on the R2R VLN benchmark by using synthetic trajectories from a prompted language model and domain transfer where a policy learned on one simulated environment (ALFRED) is transferred to another (more realistic) environment and combining both vision- and language-based representations. |
Mismatch between Multi-turn Dialogue and its Evaluation Metric in Dialogue State Tracking (2022.acl-short)
Copied to clipboard
| Challenge: | Existing evaluation metrics for dialog state tracking are limited for belief states accumulated as dialog proceeds . relative slot accuracy allows intuitive evaluation by assigning relative scores according to the turn of each dialog . |
| Approach: | They propose to use relative slot accuracy to complement existing evaluation metrics . joint goal accuracy and slot accuracy are used to evaluate accumulated belief states . |
| Outcome: | The proposed metrics focus on "penalizing states that fail to predict," not "reward for well-predicted states" the proposed metrics do not depend on the number of predefined slots, and allow intuitive evaluation . |
Aspect-Oriented Summarization for Psychiatric Short-Term Readmission Prediction (2025.emnlp-main)
Copied to clipboard
WonJin Yoon, Boyu Ren, Spencer Thomas, Chanhwi Kim, Guergana K Savova, Mei-Hua Hall, Timothy A. Miller
| Challenge: | Recent advances in large language models have enabled the automated processing of lengthy documents even without supervised training on a task-specific dataset. |
| Approach: | They propose a method for processing the summaries of long documents using different aspect-oriented prompts and integrate the information signals from these different prompts for supervised training of transformer models. |
| Outcome: | The proposed method improves on a high-impact task predicting readmissions from a psychiatric discharge using real-world data from four hospitals. |
Joint Multimodal Preference Optimization for Fine-Grained Visual-Textual Alignment (2026.findings-eacl)
Copied to clipboard
| Challenge: | Recent research has focused on addressing multimodal hallucinations in Large Vision-Language Models (LVLMs) however, these methods lack fine-grained visual contrast mechanisms and rely on single-margin optimization. |
| Approach: | They propose a framework that integrates text-conditioned preference loss with visual ranking-based objective. |
| Outcome: | The proposed framework improves cross-modal alignment and fine-grained visual grounding. |
Feature Difference Makes Sense: A medical image captioning model exploiting feature difference and tag information (2020.acl-srw)
Copied to clipboard
| Challenge: | Existing methods for medical image captioning are limited and lack diversity . current methods do not generalize well when applied to unfamiliar images . |
| Approach: | They propose a feature difference and tag information combined long short-term memory model for chest x-ray report generation. |
| Outcome: | The proposed model outperforms existing models in chest x-ray report generation. |
BlendX: Complex Multi-Intent Detection with Blended Patterns (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing datasets such as MixATIS and MixSNIPS have limitations in their formulation. |
| Approach: | They propose a set of multi-intent detection datasets that feature more diverse patterns than their predecessors. |
| Outcome: | The proposed datasets feature more diverse patterns than their predecessors and are more complex and diverse than existing datasets. |
Can You Learn Semantics Through Next-Word Prediction? The Case of Entailment (2024.findings-acl)
Copied to clipboard
| Challenge: | et al. argued that sentence co-occurrence probabilities should reflect entailment . but it is unclear whether probabilities predicted by neural LMs encode enanglement based on their theory . |
| Approach: | They propose a test that decodes entailment relations between natural sentences . they argue that the test that predicts a flipped test does not account for redundancy . |
| Outcome: | The proposed test can decode entailment relations between natural sentences, but not perfectly. |
Information-Theoretic Text Hallucination Reduction for Video-grounded Dialogue (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing video-grounded dialogue systems suffer from text hallucination problem due to learning spurious correlations from the fact that answer sentences in the dataset usually include the words of input texts. |
| Approach: | They propose to decode an answer sentence to a question using video and dialogue contexts. |
| Outcome: | The proposed framework shows that it generates adequate conversational responses to the queries of humans while following up on video and dialogue context. |
AligNART: Non-autoregressive Neural Machine Translation by Jointly Learning to Estimate Alignment and Translate (2021.emnlp-main)
Copied to clipboard
| Challenge: | Non-autoregressive neural machine translation models suffer from the multi-modality problem . aligNART leverages full alignment information to explicitly reduce the modality of the target distribution . |
| Approach: | They propose an alignment decomposition method which explicitly reduces the modality of the target distribution. |
| Outcome: | The proposed model outperforms previous models that focus on modality reduction on two translation tasks. |
Global Reward to Local Rewards: Multimodal-Guided Decomposition for Improving Dialogue Agents (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for asynchronous dialogue agents only use a single global score at the end of the session. |
| Approach: | They propose a method for aligning an LLM-based dialogue agent for long-term social dialogue . they use local implicit feedback to decompose a human-provided global Explicit reward . |
| Outcome: | The proposed approach improves the turn-level utterance generation across conversational metrics compared to baseline methods. |
DEBATE: Devil’s Advocate-Based Assessment and Text Evaluation (2024.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for evaluating the quality of machine-generated texts have a relatively low correlation with human performance. |
| Approach: | They propose an NLG evaluation framework based on multi-agent scoring system augmented with a concept of Devil’s Advocate. |
| Outcome: | The proposed evaluation framework outperforms the previous state-of-the-art methods in two meta-evaluation benchmarks in NLG evaluation, SummEval and TopicalChat. |
PVP: An Image Dataset for Personalized Visual Persuasion with Persuasion Strategies, Viewer Characteristics, and Persuasiveness Ratings (2025.acl-long)
Copied to clipboard
| Challenge: | Visual persuasion uses visual elements to influence cognition and behaviors . lack of comprehensive data sets connect persuasiveness of images with personal information . |
| Approach: | They propose to use a dataset to connect persuasiveness with personal information . they find psychological characteristics enhance the generation and evaluation of persuasive images . |
| Outcome: | The proposed dataset provides persuasiveness scores of images evaluated by human annotators along with demographic and psychological characteristics. |
HEAR: Hearing Enhanced Audio Response for Video-grounded Dialogue (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing systems are competent only to incorporate information in the video and text and tend to struggle in extracting the necessary information from the audio when generating appropriate responses to the question. |
| Approach: | They propose to perform sensible listening by selectively attending to audio whenever the question requires it. |
| Outcome: | The proposed framework enhances the accuracy and audibility of VGD systems in a model-agnostic manner. |
Deep Latent Variable Models of Natural Language (D18-3)
Copied to clipboard
| Challenge: | In this tutorial, we will discuss the challenges of applying neural variational inference to NLP problems. |
| Approach: | The tutorial will cover deep latent variable models in the case where exact inference over the latent variables is tractable. |
| Outcome: | The proposed tutorial will cover deep latent variable models in the case where inference cannot be performed tractably and when it is not . |
FIZZ: Factual Inconsistency Detection by Zoom-in Summary and Zoom-out Document (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for evaluating factual consistency in abstractive summarization systems have significant limitations, especially on refinement and interpretability. |
| Approach: | They propose a method for detecting summary factual inconsistency based on fine-grained atomic facts decomposition and adaptive granularity expansion. |
| Outcome: | The proposed method outperforms existing systems on the AGGREFACT benchmark dataset and achieves state-of-the-art performance. |
K-MHaS: A Multi-label Hate Speech Detection Dataset in Korean Online News Comment (2022.coling-1)
Copied to clipboard
| Challenge: | Online hate speech detection resources in other languages are limited. |
| Approach: | They introduce a new dataset for hate speech detection that handles Korean language patterns. |
| Outcome: | The proposed dataset outperforms existing datasets in Korean language classifications. |
Fine-tuning CLIP Text Encoders with Two-step Paraphrasing (2024.findings-eacl)
Copied to clipboard
| Challenge: | Contrastive language-image pre-training models have demonstrated considerable success across various vision-language tasks, such as text-to-image retrieval. |
| Approach: | They propose a fine-tuning approach to enhance the representations of CLIP models for paraphrases by leveraging large language models. |
| Outcome: | The proposed model improves on baseline models across paraphrased retrieval, visual genome relation and attribution, and seven semantic textual similarity tasks. |
The BiGGen Bench: A Principled Benchmark for Fine-grained Evaluation of Language Models with Language Models (2025.naacl-long)
Copied to clipboard
Seungone Kim, Juyoung Suk, Ji Yong Cho, Shayne Longpre, Chaeeun Kim, Dongkeun Yoon, Guijin Son, Yejin Cho, Sheikh Shafayat, Jinheon Baek, Sue Hyun Park, Hyeonbin Hwang, Jinkyung Jo, Hyowon Cho, Haebin Shin, Seongyun Lee, Hanseok Oh, Noah Lee, Namgyu Ho, Se June Joo, Miyoung Ko, Yoonjoo Lee, Hyungjoo Chae, Jamin Shin, Joel Jang, Seonghyeon Ye, Bill Yuchen Lin, Sean Welleck, Graham Neubig, Moontae Lee, Kyungjae Lee, Minjoon Seo
| Challenge: | a recent study evaluated language models using abstract evaluation criteria that lack the flexibility and granularity of human assessment. |
| Approach: | They propose a benchmark to evaluate nine distinct language models' capabilities . they use instance-specific evaluation criteria to mirror human evaluation . |
| Outcome: | The proposed benchmark evaluates nine distinct capabilities of language models across 77 tasks. |
Interpretation of NLP models through input marginalization (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to interpret NLP predictions replace each token with a predefined value, resulting in misleading interpretations. |
| Approach: | They propose to marginalize each token out of the training data distribution to demystify the "black box" property of deep neural networks for natural language processing. |
| Outcome: | The proposed method marginalizes each token out of the training data distribution. |
CHAMP: A Competition-level Dataset for Fine-Grained Analyses of LLMs’ Mathematical Reasoning Capabilities (2024.findings-acl)
Copied to clipboard
| Challenge: | Recent large language models have shown indications of mathematical reasoning ability on competition-level problems. |
| Approach: | They propose a benchmark dataset to enable such analyses using large language models. |
| Outcome: | The proposed model performs better with concepts and hints than with the best model, but it is difficult to verify. |
Effective Crowdsourcing of Multiple Tasks for Comprehensive Knowledge Extraction (2020.lrec-1)
Copied to clipboard
Sangha Nam, Minho Lee, Donghwan Kim, Kijong Han, Kuntae Kim, Sooji Yoon, Eun-kyung Kim, Key-Sun Choi
| Challenge: | Existing studies on information extraction from unstructured texts lack a coherent evaluation of all tasks. |
| Approach: | They propose to use crowdsourcing data to develop a Korean information extraction initiative point . they propose to train and evaluate four Korean information extracting tasks using a state-of-the-art model . |
| Outcome: | The proposed model will be used to evaluate four Korean information extraction tasks using crowdsourcing data. |
C2: Scalable Auto-Feedback for LLM-based Chart Generation (2025.naacl-long)
Copied to clipboard
Woosung Koh, Janghan Yoon, MinHyung Lee, Youngjin Song, Jaegwan Cho, Jaehyun Kang, Taehyeon Kim, Se-Young Yun, Youngjae Yu, Bongshin Lee
| Challenge: | generating high-quality charts with Large Language Models presents significant challenges due to limited data and the high cost of curation. |
| Approach: | They propose a referencefree automatic feedback generator to generate high-quality charts with Large Language Models. |
| Outcome: | The proposed framework outperforms baselines and shows that it significantly improves data diversity. |
LLM as a Risk Manager: LLM Semantic Filtering for Lead–Lag Trading in Prediction Markets (2026.acl-industry)
Copied to clipboard
Sumin Kim, Minjae Kim, Jihoon Kwon, Yoon Kim, Oscar Levy, Alejandro Lopez-Lira, Yongjae Lee, Chanyeol Choi
| Challenge: | Prediction markets provide a unique setting where event-level time series are directly tied to natural-language descriptions, yet discovering robust lead–lag relationships remains challenging due to spurious statistical correlations. |
| Approach: | They propose a statistical stage that uses Granger causality to identify candidate leader–follower pairs from market-implied probability time series and an LLM-based semantic stage that re-ranks these candidates by assessing whether the proposed direction admits a plausible economic transmission mechanism. |
| Outcome: | The proposed approach consistently outperforms the statistical baseline on Kalshi Economics markets. |
Carpe diem: On the Evaluation of World Knowledge in Lifelong Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Current language models are trained on static data, implying that the encoded knowledge could go wrong as time passes. |
| Approach: | They propose a temporally evolving question-answering benchmark for language models . they use Wikipedia databases to test language models for dynamic knowledge in ever-changing world . |
| Outcome: | The proposed task aims to model the evolution-adaptability of language models in the real world. |
Efficient Latent Variable Modeling for Knowledge-Grounded Dialogue Generation (2023.findings-emnlp)
Copied to clipboard
Gunsoo Han, Daejin Jo, Daniel Nam, Eunseop Yoon, Taehwan Kwon, Seungeun Rho, Kyoung-Woon On, Chang Yoo, Sungwoong Kim
| Challenge: | Existing knowledge-grounded dialogue generation algorithms require annotated knowledge to generate a response grounded on the retrieved knowledge. |
| Approach: | They propose an efficient algorithm for latent variable modeling that leverages large amount of dialogue data. |
| Outcome: | The proposed algorithm outperforms the supervised learning algorithm on knowledge-grounded dialogue datasets while maintaining efficiency and scalability. |
CCQA: Generating Question from Solution Can Improve Inference-Time Reasoning in SLMs (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in large language models (LLMs) have yielded remarkable performance across a wide range of tasks, including machine translation, code generation, sentiment analysis, and reasoning. |
| Approach: | They propose a new reasoning method that generates a question from each reasoning path and answer, evaluates each by its similarity to the original question, and selects the candidate solution with the highest similarity score as the final answer. |
| Outcome: | The proposed method outperforms existing state-of-the-art methods on mathematical and commonsense reasoning benchmarks and establishes a new practical baseline for efficient reasoning in SLMs. |
Towards standardizing Korean Grammatical Error Correction: Datasets and Annotation (2023.acl-long)
Copied to clipboard
| Challenge: | Despite the growing number of Korean learners, little research has been conducted on Korean grammatical error correction (GEC) despite the difficulties of the Korean language, there is no evaluation benchmark for Korean GEC. |
| Approach: | They propose to use Korean grammar error correction datasets to train a machine learning model that can automatically annotate Korean errors from parallel corpora. |
| Outcome: | The proposed model outperforms the currently used statistical Korean GEC system on a wider range of error types. |
Towards Personalized Conversational Sales Agents: Contextual User Profiling for Strategic Action (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Conversational Recommender Systems (CRSs) aim to engage users in dialogue to provide tailored recommendations. |
| Approach: | They propose a task that integrates preference elicitation, recommendation and persuasion . they propose CSALES, a conversational sales agent that infers contextual user profiles . |
| Outcome: | The proposed task integrates preference elicitation, recommendation, and persuasion within a conversational framework. |
LangBridge: Multilingual Reasoning Without Multilingual Supervision (2024.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to adapt language models for multilingual reasoning tasks require multilingual supervision. |
| Approach: | They propose a zero-shot approach to adapt language models for multilingual reasoning tasks without multilingual supervision by bridging two models by introducing minimal trainable parameters between them. |
| Outcome: | The proposed approach significantly improves multilingual reasoning capabilities on low-resource languages. |
Instruction Tuning with and without Context: Behavioral Shifts and Downstream Impact (2026.eacl-long)
Copied to clipboard
Hyunji Lee, Seunghyun Yoon, Yunjae Won, Hanseok Oh, Geewook Kim, Trung Bui, Franck Dernoncourt, Elias Stengel-Eskin, Mohit Bansal, Minjoon Seo
| Challenge: | Prior work on instruction tuning datasets combined these data types without examining their distinct effects. |
| Approach: | They investigate how training LLMs with or without context affects model behavior and performance . they find that using context-augmented data as the backbone for vision-language models reduces hallucination . |
| Outcome: | The proposed training with context-augmented data reduces hallucination and improves grounding in the visual domain. |
GUI Agents: A Survey (2025.findings-acl)
Copied to clipboard
Dang Nguyen, Jian Chen, Yu Wang, Gang Wu, Namyong Park, Zhengmian Hu, Hanjia Lyu, Junda Wu, Ryan Aponte, Yu Xia, Xintong Li, Jing Shi, Hongjie Chen, Viet Dac Lai, Zhouhang Xie, Sungchul Kim, Ruiyi Zhang, Tong Yu, Mehrab Tanjim, Nesreen K. Ahmed, Puneet Mathur, Seunghyun Yoon, Lina Yao, Branislav Kveton, Jihyung Kil, Thien Huu Nguyen, Trung Bui, Tianyi Zhou, Ryan A. Rossi, Franck Dernoncourt
| Challenge: | Large Foundation Models (LFMs) have transformed the landscape of AI research and day-to-day life. |
| Approach: | They propose a framework that delineates GUI agents' perception, reasoning, planning, and acting capabilities. |
| Outcome: | The proposed framework delineates their perception, reasoning, planning, and acting capabilities. |
Hierarchical Phrase-Based Sequence-to-Sequence Learning (2022.emnlp-main)
Copied to clipboard
| Challenge: | a neural transducer that incorporates hierarchical phrases as a source of inductive bias during training and as explicit constraints during inference is described. |
| Approach: | They propose a neural transducer that incorporates hierarchical phrases as a source of inductive bias during training and as explicit constraints during inference. |
| Outcome: | The proposed model performs well on small scale machine translation benchmarks. |
reWordBench: Benchmarking and Improving the Robustness of Reward Models with Transformed Inputs (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing reward models have a high performance on benchmarks, but performance degradation is often due to overfitting. |
| Approach: | They propose to explicitly train reward models to assign similar scores to paraphrases to improve their robustness. |
| Outcome: | The proposed model reduces degradation by half for the Chat Hard subset in RewardBench. |
Explain-then-translate: an analysis on improving program translation with self-generated explanations (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Using self-generated natural language explanations improves zero-shot performance by 12% on average. |
| Approach: | They propose to use self-generated natural language explanations as an intermediate step for code-to-code translation with language models. |
| Outcome: | The proposed approach improves zero-shot performance by 12% on average . the proposed approach is not evaluated on a broader set of languages including low-resource languages. |
What Do Language Models Hear? Probing for Auditory Representations in Language Models (2024.acl-long)
Copied to clipboard
| Challenge: | a linear probe is used to retrieve the correct text representation of an object given a snippet of audio related to that object. |
| Approach: | They develop a linear probe that retrieves the correct text representation of an object . they then test the probe's generalization to objects that were not seen during training . |
| Outcome: | The proposed model generalizes to objects that were not seen during training, the study finds . the model can learn representations of perceptual concepts that plausibly mirror the grounded representations . |
Auditing LLM Responses to Harmful Stereotypes Targeting Mental Health Groups (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) can exhibit imbalanced biases against vulnerable groups, but how they rationalize stereotypes and rights restrictions targeting mental health entities remains underexplored. |
| Approach: | They audit a suite of open-weight LLMs on stereotype-justification prompts tied to mental health identities. |
| Outcome: | The proposed models endorse harmful stereotypes when explicitly asked to justify them, with endorsement varying across model families, versions, and mental health conditions. |
TelAgentBench: A Multi-faceted Benchmark for Evaluating LLM-based Agents in Telecommunications (2025.emnlp-industry)
Copied to clipboard
Sunwoo Lee, Daseong Jang, Dhammiko Arya, Gyoung-eun Han, Injee Song, SaeRom Kim, Sangjin Kim, Seojin Lee, Seokyoung Hong, Sereimony Sek, Seung-Mo Cho, Sohee Park, Sungbin Yoon, Wonbeom Jang, Eric Davis
| Challenge: | Large Language Models (LLMs) are becoming powerful agentic systems . generic benchmarks fail to assess realistic, non-English performance . |
| Approach: | They propose to evaluate five core agentic capabilities: Reasoning, Planning, Action (tool-use), Retrieval-Augmented Generation, and Instruction Following. |
| Outcome: | The evaluations reveal significant performance disparities between models that employ explicit reasoning and those that do not. |
Optimizing Language Augmentation for Multilingual Large Language Models: A Case Study on Korean (2024.lrec-main)
Copied to clipboard
ChangSu Choi, Yongbin Jeong, Seoyoon Park, Inho Won, HyeonSeok Lim, SangMin Kim, Yejee Kang, Chanhyuk Yoon, Jaewan Park, Yiseul Lee, HyeJin Lee, Younggyun Hahm, Hansaem Kim, KyungTae Lim
| Challenge: | Large language models (LLMs) use pretraining to predict the subsequent word, but less-resourced languages are being overlooked. |
| Approach: | They propose to expand the MLLM vocabularies to enhance expressiveness and use bilingual data for pretraining to align the high- and less-resourced languages. |
| Outcome: | The proposed model outperforms existing models in qualitative analyses compared to Korean monolingual models. |
Ask Optimal Questions: Aligning Large Language Models with Retriever’s Preference in Conversation (2025.findings-naacl)
Copied to clipboard
| Challenge: | Existing methods to perform conversational search are sub-optimal due to the limited ability to incorporate signals from the retrieval results. |
| Approach: | They propose to optimize a language model for reformulating search queries in line with retrievers’ preferences by combining a large-scale dataset with Retrievers’ Feedback. |
| Outcome: | The proposed framework outperforms existing methods on two benchmarks and surpasses the state-of-the-art methods. |
Gradient Ascent Post-training Enhances Language Model Generalization (2023.acl-short)
Copied to clipboard
| Challenge: | Recent studies show that language models can perform diverse downstream NLP tasks in a zero-shot manner without any gradient updates. |
| Approach: | They show that updating pretrained LMs with Gradient Ascent Post-training enhances zero-shot generalization capabilities without any task-specific fine-tuning. |
| Outcome: | The proposed method can be comparable to 2-3x times larger LMs across 12 different NLP tasks. |
ListT5: Listwise Reranking with Fusion-in-Decoder Improves Zero-shot Retrieval (2024.acl-long)
Copied to clipboard
| Challenge: | Existing listwise reranking models rely on pointwise sizing of each passage . Until now, listwise models lack the ability to compare between passages at inference time . |
| Approach: | They propose a listwise reranking approach based on Fusion-in-Decoder that handles multiple candidate passages at train and inference time. |
| Outcome: | The proposed model outperforms the state-of-the-art RankT5 model on the BEIR benchmark for zero-shot retrieval task with a notable +1.3 gain in the average NDCG@10 score. |
SSMix: Saliency-Based Span Mixup for Text Classification (2021.findings-acl)
Copied to clipboard
| Challenge: | SSMix synthesizes a sentence while preserving the locality of two original texts by span-based mixing and keeping more tokens related to the prediction relying on saliency information. |
| Approach: | They propose a new method where the operation is performed on input text rather than on hidden vectors like previous approaches. |
| Outcome: | The proposed method outperforms hidden-level mixup methods on a wide range of text classification benchmarks including textual entailment, sentiment classification, and questiontype classification. |
Leveraging Large Language Models for Learning Complex Legal Concepts through Storytelling (2024.acl-long)
Copied to clipboard
Hang Jiang, Xiajie Zhang, Robert Mahari, Daniel Kessler, Eric Ma, Tal August, Irene Li, Alex Pentland, Yoon Kim, Deb Roy, Jad Kabbara
| Challenge: | a novel application of large language models (LLMs) to legal education helps non-experts learn complex legal concepts . authors find storytelling helps nonexperts understand complex legal terms and concepts compared to definitions . |
| Approach: | They propose a novel application of large language models to legal education . they use LLMs to generate legal stories explaining complex legal concepts . |
| Outcome: | The proposed method improves comprehension and interest among non-native speakers compared to definitions . the novel method also shows that non-experts retain more stories . |
Implicit Representations of Grammaticality in Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Pretrained language models generate grammatically well-formed text and discriminate well between grammatical and ungrammatically sentences in tightly controlled minimal pairs. |
| Approach: | They propose a method to train pretrained LMs for representations of grammaticality by applying perturbations to a naturalistic text corpus. |
| Outcome: | The proposed model outperforms probability-based models on human-curated grammaticality judgment benchmarks and performs worse than string probabilities on plausibility benchmarks. |
RA-RRG: Multimodal Retrieval-Augmented Radiology Report Generation with Key Phrase Extraction (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing MLLMs are computationally expensive and may produce hallucinated content . RA-RRG uses large language models to generate radiology reports . |
| Approach: | They propose a retrieval-augmented RRG framework that combines multimodal retrieval with large language models to generate radiology reports. |
| Outcome: | RA-RRG uses large language models to generate radiology reports . it suppresses hallucinations while maintaining strong report generation performance . |
DUSK: Do Not Unlearn Shared Knowledge (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent work suggests that machine learning models are indistinguishable from models trained on retain sets. |
| Approach: | They propose a benchmark to evaluate machine unlearning under realistic knowledge overlap . they construct documents containing both shared and unique knowledge . |
| Outcome: | The proposed model is indistinguishable from a model retrained on the retain set while only forget-specific content is removed. |
AutoAnoEval: Semantic-Aware Model Selection via Tree-Guided LLM Reasoning for Tabular Anomaly Detection (2026.findings-eacl)
Copied to clipboard
Suhee Yoon, Sanghyu Yoon, Ye Seul Sim, Seungdong Yoa, Dongmin Kim, Soonyoung Lee, Hankook Lee, Woohyung Lim
| Challenge: | Existing approaches to tabular anomaly detection fail to reflect domain specific nature of real-world anomalies. |
| Approach: | They propose a framework that constructs pseudo-evaluation sets with semantically grounded synthetic anomalies. |
| Outcome: | The proposed framework generates pseudo-evaluation sets with semantically grounded synthetic anomalies. |
Inducing and Using Alignments for Transition-based AMR Parsing (2022.naacl-main)
Copied to clipboard
Andrew Drozdov, Jiawei Zhou, Radu Florian, Andrew McCallum, Tahira Naseem, Yoon Kim, Ramón Astudillo
| Challenge: | Abstract Meaning Representation parsers rely on node-to-word alignments, but lack the complexity of the pipeline. |
| Approach: | They propose a neural aligner for abstract meaning representation that learns node-to-word alignments without relying on pipelines. |
| Outcome: | The proposed approach improves accuracy and generalization from AMR2.0 to AMR3.0 corpora. |
Still Between Us? Evaluating and Improving Voice Assistant Robustness to Third-Party Interruptions (2026.acl-long)
Copied to clipboard
| Challenge: | Recent Spoken Language Models lack the capability to discern Third-Party Interruptions (TPI) from the primary user’s ongoing flow, leaving them vulnerable to contextual failures. |
| Approach: | They propose a dataset with speaker-aware hard negatives to enforce acoustic cue prioritization for interruption handling and a framework to measure the interruption-handling strategy and precise speaker discrimination in deceptive contexts. |
| Outcome: | The proposed framework mitigates semantic shortcut learning while neglecting acoustic signals essential for discerning speaker changes. |
Zero-shot Multimodal Document Retrieval via Cross-modal Question Generation (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing multimodal large language models struggle when faced with unseen domains or languages. |
| Approach: | They propose a framework that leverages the broad knowledge of an MLLM to generate cross-modal pre-questions (preQs) before retrieval. |
| Outcome: | Experiments show that PREMIR outperforms existing retrievers on out-of-distribution benchmarks, including closed-domain and multilingual settings, outperforming strong baselines across all metrics. |
Language-Grounded Multi-Domain Image Translation via Semantic Difference Guidance (2026.eacl-long)
Copied to clipboard
| Challenge: | Existing methods for image-to-image translation lack structural integrity and attribute-specific control . Existing approaches lack semantics and provide fine-grained, attribute-based control compared to GAN-based methods . |
| Approach: | They propose a language-grounded attribute-controllable translation framework that grounds semantic differences into corresponding visual transformations while preserving unrelated structural and semantic content. |
| Outcome: | Experiments on CelebA(Dialog) and BDD100K show that LACE achieves high visual fidelity, structural preservation, and interpretable domain-specific control, surpassing baselines. |
Lookback Lens: Detecting and Mitigating Contextual Hallucinations in Large Language Models Using Only Attention Maps (2024.emnlp-main)
Copied to clipboard
| Challenge: | Despite the utility and impressive capabilities of large language models, their tendency to generate hallucinations presents a significant challenge in their deployment. |
| Approach: | They propose a simple hallucination detection model based on the ratio of attention weights on the context versus newly generated tokens. |
| Outcome: | The proposed model reduces the amount of hallucinations by 9.6% in a summarization task. |
Paraphrase and Solve: Exploring and Exploiting the Impact of Surface Form on Mathematical Reasoning in Large Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Despite the impressive performance of large-scale language models, their ability to reason through complex problems remains a bottleneck. |
| Approach: | They propose a method which diversifies reasoning paths from specific surface forms of the problem to improve mathematical reasoning performance. |
| Outcome: | The proposed approach improves mathematical reasoning performance over vanilla self-consistency, especially for problems initially deemed unsolvable. |
Unsupervised Discontinuous Constituency Parsing with Mildly Context-Sensitive Grammars (2023.acl-long)
Copied to clipboard
| Challenge: | a recent study shows that context-free grammars are not natural for modeling discontinuous language phenomena such as extrapositions and cross-serial dependencies. |
| Approach: | They propose a grammar induction approach with mildly context-sensitive grammars for unsupervised discontinuous parsing. |
| Outcome: | Experiments on German and Dutch show that the proposed grammar induction method is beneficial for unsupervised parsing. |
Unsupervised Recurrent Neural Network Grammars (N19-1)
Copied to clipboard
| Challenge: | RNNGs model syntax and structure by incrementally generating a syntax tree and sentence in a top-down, left-to-right order. |
| Approach: | They explore unsupervised learning of recurrent neural network grammars for language modeling and grammar induction. |
| Outcome: | The proposed model outperforms standard sequential language models and improves parsing performance. |
Probing for Incremental Parse States in Autoregressive Language Models (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Existing work on autoregressive language models shows impressive command of syntax . implicit incremental syntactic inferences underlie next-word predictions . |
| Approach: | They propose a probe for extracting incomplete syntactic structure from autoregressive language models. |
| Outcome: | The proposed probes can predict model preferences on ambiguous sentence prefixes and causally intervene on model representations and steer model behavior. |