RE-TRAINING

Fantastic Expressions and Where to Find Them: Chinese Simile Generation with Multiple Constraints (2023.acl-long)

Challenge:	Existing work focuses on generating sentences satisfying pre-specified attributes such as topic and sentiment, yet suffers from increases in storage and inference time.
Approach:	They propose a method that uses a pre-trained continuous vector to generate a fixed pre-trainable language model to satisfy a specified attribute.
Outcome:	The proposed model can achieve improvements on eleven attribute-specific generation tasks with 0.08% extra training parameters.

Copied to clipboard

Kexin Yang, Dayiheng Liu, Wenqiang Lei, Baosong Yang, Xiangpeng Wei, Zhengyuan Liu, Jun Xie

Information Retrieval

Challenge:	Existing attempts to generate similes as context-free tasks are not suitable for simile generation . however, simile generated under such settings might be undesirable, we argue .
Approach:	They propose a model to generate a simile with multiple simile elements . they propose to use a vehicle retrieval module to obtain the explicable comparison .
Outcome:	The proposed model can generate a simile with multiple simile elements, e.g., context and vehicle.

Compute-Efficient Churn Reduction for Conversational Agents (2023.emnlp-industry)

Copied to clipboard

Christopher Hidey, Sarthak Sarthak

Reliable and Interpretable Drift Detection in Streams of Short Texts (2023.acl-industry)

Challenge:	churn occurs when retraining models yields different predictions despite using the same data and hyper-parameters.
Approach:	They propose a method that pairs semantic parses based on their “function call signature” and encourages similarity through an additional loss based upon Jensen-Shannon Divergence.
Outcome:	The proposed method improves in academic, noisy, and industry settings.

Copied to clipboard

Ella Rabinovich, Matan Vetzler, Samuel Ackerman, Ateret Anaby Tavor

A Versatile Adaptive Curriculum Learning Framework for Task-oriented Dialogue Policy Learning (2022.findings-naacl)

Challenge:	Data drift is a key factor leading to model performance degradation over time.
Approach:	They propose a framework for reliable model-agnostic change-point detection and interpretation in large task-oriented dialog systems.
Outcome:	The proposed framework is effective in multiple customer deployments.

Copied to clipboard

Yang Zhao, Hua Qin, Wang Zhenyu, Changxi Zhu, Shihan Wang

Weakly supervised hierarchical multi-task classification of customer questions (2023.acl-industry)

Challenge:	Existing training paradigms for dialogue policy learning with brute-force random sampling are expensive and lack reliable evaluation of difficulty scores.
Approach:	They propose a flexible adaptive curriculum learning framework that integrates curriculum learning with a generic global curriculum.
Outcome:	The proposed framework improves learning performance and efficiency on three public dialogue datasets.

Copied to clipboard

Jitenkumar Rana, Promod Yenigalla, Chetan Aggarwal, Sandeep Sricharan Mukku, Manan Soni, Rashmi Patange

Macro-Average: Rare Types Are Important Too (2021.naacl-main)

Challenge:	Identifying granular and actionable topics from customer questions helps improve the overall customer experience.
Approach:	They propose a weakly supervised Hierarchical Multi-task Classification Framework to identify granular topics from customer questions . a clustering based taxonomy creation and data labeling module is used to create taxonomies and labelled data with minimal supervision.
Outcome:	The proposed model achieves 13% better accuracy over single-task classification frameworks . it can adapt to constantly evolving taxonomy without need of re-training .

Copied to clipboard

Thamme Gowda, Weiqiu You, Constantine Lignos, Jonathan May

Multilinguality

Challenge:	MT metrics trained on segment-level human judgments are inherently non-transparent and reflect undesirable biases.
Approach:	They propose to use a type-based classifier metric to evaluate machine translation and compare it with a supervised and unsupervised one.
Outcome:	The proposed model outperforms other models in indicating cross-lingual information retrieval task performance and shows that it can be used to compare supervised and unsupervised neural machine translation.

Incremental Few-shot Text Classification with Multi-round New Classes: Formulation, Dataset and System (2021.naacl-main)

Copied to clipboard

Congying Xia, Wenpeng Yin, Yihao Feng, Philip Yu

Enhancing Self-Consistency and Performance of Pre-Trained Language Models through Natural Language Inference (2022.emnlp-main)

Challenge:	Text classification is usually studied by labeling texts with relevant categories from a predefined set.
Approach:	They propose a task where a system incrementally handles multiple rounds of new classes . they propose two entailment approaches, ENTAILMENT and HYBRID, which show promise .
Outcome:	The proposed task is based on a few-shot text classification task in the NLP domain.

Copied to clipboard

Eric Mitchell, Joseph Noh, Siyan Li, Will Armstrong, Ananth Agarwal, Patrick Liu, Chelsea Finn, Christopher Manning

Towards Continuous Dialogue Corpus Creation: writing to corpus and generating from it (L18-1)

Challenge:	Existing frameworks for boosting consistency and accuracy of pre-trained NLP models without fine-tuning or re-training are lacking.
Approach:	They propose a framework for boosting the consistency and accuracy of pre-trained NLP models using pre-trainer natural language inference models without fine-tuning or re-training.
Outcome:	The proposed framework boosts consistency and accuracy of pre-trained NLP models using pre-train natural language inference models without fine-tuning or re-training.

Copied to clipboard

Andrei Malchanau, Volha Petukhova, Harry Bunt

GFST: Gender-Filtered Self-Training for More Accurate Gender in Translation (2021.emnlp-main)

Challenge:	Existing methods to create dialogue corpora annotated with interoperable semantic information are based on ISO standard data models and tools.
Approach:	They propose to use a corpus as a shared repository for analysis and modelling of interactive dialogue behaviour and for implementation, integration and evaluation of dialogue system components.
Outcome:	The proposed method is applied to the design of two multimodal interactive applications - the Virtual Negotiation Coach and the Virtual Debate Coach.

Copied to clipboard

Prafulla Kumar Choubey, Anna Currey, Prashant Mathur, Georgiana Dinu

Multilinguality

Challenge:	Recent studies have focused on gender bias in neural machine translation (NMT) incorrectly gendered translations can reflect or amplify social biases.
Approach:	They propose to use a monolingual corpus to generate gender-specific pseudo-parallel corpora and filter them to improve gender translation accuracy.
Outcome:	The proposed approach improves gender accuracy without damaging generic quality on translations from English into five languages.

PEEB: Part-based Image Classifiers with an Explainable and Editable Language Bottleneck (2024.findings-naacl)

Copied to clipboard

Thang Pham, Peijie Chen, Tin Nguyen, Seunghyun Yoon, Trung Bui, Anh Nguyen

ReFACT: Updating Text-to-Image Models by Editing the Text Encoder (2024.naacl-long)

Challenge:	CLIP-based classifiers rely on the prompt containing a class name that is known to the text encoder and perform poorly on new classes or the classes whose names rarely appear on the Internet.
Approach:	They propose to use a set of text descriptors to express a class name into a textual descriptable and match the embeddings of the detected parts to their textual ones to compute a logit score.
Outcome:	The proposed classifier outperforms CLIP-based classifiers on zero-shot and supervised learning settings by 88.80% and 92.20% accuracy on CUB-200 and Stanford Dogs-120.

Copied to clipboard

Dana Arad, Hadas Orgad, Yonatan Belinkov

Multimodality

Challenge:	Text-to-image generative models encode factual associations that can quickly become outdated, diminishing their utility for end-users.
Approach:	They propose a method for editing factual associations in text-to-image models without retraining or explicit input from end-users.
Outcome:	The proposed method improves generalization and preservation of unrelated concepts on an existing dataset and compares with other methods.

A Two-phase Prototypical Network Model for Incremental Few-shot Relation Classification (2020.coling-main)

Copied to clipboard

Haopeng Ren, Yi Cai, Xiaofeng Chen, Guohua Wang, Qing Li

Information Extraction & Text Mining

Challenge:	Existing supervised and distantly supervised RC models ignore the emergence of novel relations in open environment.
Approach:	They propose a two-phase prototypical network with prototype attention alignment and triplet loss to dynamically recognize the novel relations with a few support instances without catastrophic forgetting.
Outcome:	Experiments show that the proposed model performs better on deep learning and few-shot learning . it can recognize the novel relations with a few support instances without catastrophic forgetting .

Defeating Cerberus: Privacy-Leakage Mitigation in Vision Language Models (2026.findings-eacl)

Copied to clipboard

Boyang Zhang, Istemi Ekin Akkus, Ruichuan Chen, Alice Dethise, Klaus Satzke, Ivica Rimac, Yang Zhang

Multimodality

Challenge:	Existing models that process multiple modalities of data have been used for multimodal tasks, but their advanced capabilities raise privacy concerns.
Approach:	They propose a method to modify the model’s internal states associated with PII-related content and to reduce the risk of PI I leakage by modifying the model's internal state.
Outcome:	The proposed method achieves on average 93.3% refusal rate for various PII-related tasks with minimal impact on unrelated model performances.

DyLex: Incorporating Dynamic Lexicons into BERT for Sequence Labeling (2021.emnlp-main)

Copied to clipboard

Baojun Wang, Zhao Zhang, Kun Xu, Guang-Yuan Hao, Yuyang Zhang, Lifeng Shang, Linlin Li, Xiao Chen, Xin Jiang, Qun Liu

Training Data Augmentation for Detecting Adverse Drug Reactions in User-Generated Content (D19-1)

Challenge:	Existing approaches to integrate lexical knowledge into deep learning models are limited by large-scale dynamic lexicons.
Approach:	They propose a plug-in lexicon incorporation approach for BERT based sequence labeling tasks . they adopt word-agnostic tag embeddings to avoid re-training the representation .
Outcome:	The proposed framework achieves new SOTA even with large scale lexicons, the authors show . they adopt word-agnostic tag embeddings to avoid re-training the representation .

Copied to clipboard

Sepideh Mesbah, Jie Yang, Robert-Jan Sips, Manuel Valle Torre, Christoph Lofi, Alessandro Bozzon, Geert-Jan Houben

ConvGQR: Generative Query Reformulation for Conversational Search (2023.acl-long)

Challenge:	Existing dictionary-based, semi-supervised learning approaches are limited by the coverage and maintainability of laymen health vocabularies.
Approach:	They propose a data augmentation approach that leverages variational autoencoders to learn high-quality data distributions from a large unlabeled dataset and generate a small set of labeled training sets.
Outcome:	The proposed approach matches the performance of fully-supervised approaches while requiring only 25% of training data.

Copied to clipboard

Fengran Mo, Kelong Mao, Yutao Zhu, Yihong Wu, Kaiyu Huang, Jian-Yun Nie

Natural Language Interfaces

Challenge:	Existing methods to determine a good search query from the whole conversation context are expensive and often lead to sub-optimal results.
Approach:	They propose a framework to reformulate conversational queries based on generative pre-trained language models (PLMs) they propose generative knowledge infusion mechanism to optimize query reformulation and retrieval.
Outcome:	Extensive experiments on four conversational search datasets demonstrate the effectiveness of ConvGQR.

Adaptable and Interpretable Neural MemoryOver Symbolic Knowledge (2021.naacl-main)

Copied to clipboard

Pat Verga, Haitian Sun, Livio Baldini Soares, William Cohen

Neural Entity Recognition with Gazetteer based Fusion (2021.findings-acl)

Challenge:	Past research has shown that large neural language models encode surprising amounts of factual information, but augmenting or modifying this information requires modifying a corpus and retraining, which is computationally expensive.
Approach:	They propose a neural LM that includes an interpretable neuro-symbolic KB in the form of a "fact memory" their LM improves performance on knowledge-intensive question-answering tasks, sometimes dramatically .
Outcome:	The proposed model improves on knowledge-intensive question-answering tasks, including a 27 point increase in one setting of WebQuestionsSP over a state-of-the-art open-book model.

Copied to clipboard

Qing Sun, Parminder Bhatia

Total Recall: a Customized Continual Learning Method for Neural Semantic Parsers (2021.emnlp-main)

Challenge:	Named entity recognition systems can be applied to clinical domains where only limited data is accessible and interpretability is important.
Approach:	They propose to use auxiliary gazetteer model to fuse it with NER system . this allows for better robustness and interpretability across different clinical datasets .
Outcome:	The proposed model is data efficient and can adapt to new mentions in gazetteers without retraining.

Copied to clipboard

Zhuang Li, Lizhen Qu, Gholamreza Haffari

Lexically Constrained Neural Machine Translation with Levenshtein Transformer (2020.acl-main)

Challenge:	Existing methods for continual learning for semantic parsing fail to account for special properties of structured outputs . retraining from scratch is not feasible due to the fast growing number of tasks .
Approach:	They propose a continual learning method that uses sequential learning to learn tasks without accessing full training data from previous tasks.
Outcome:	The proposed method achieves a 3-6 times speedup compared to re-training from scratch.

Copied to clipboard

Raymond Hendy Susanto, Shamil Chollampatt, Liling Tan

Multilinguality

Challenge:	Existing approaches to incorporate lexical constraints in neural machine translation have been unsuccessful .
Approach:	They propose an algorithm that incorporates lexical constraints into neural machine translation.
Outcome:	The proposed method improves on English-German datasets without modification . it does not require any modification to the training procedure and can be easily applied at runtime with custom dictionaries.

E-LANG: Energy-Based Joint Inferencing of Super and Swift Language Models (2022.acl-long)

Copied to clipboard

Mohammad Akbari, Amin Banitalebi-Dehkordi, Yong Zhang

MockingBERT: A Method for Retroactively Adding Resilience to NLP Models (2022.coling-1)

Challenge:	E-LANG is an efficient method for building large and highly capable language models . existing methods are only applicable to encoder-only backbones and classification tasks .
Approach:	They propose an efficient dynamic inference approach which distributes inference between large accurate Super-models and light-weight Swift models.
Outcome:	The proposed method outperforms existing methods on GLUE, SuperGLUE and WMT with 3.3X computation speed and 2.9X computation cost.

Copied to clipboard

Jan Jezabek, Akash Singh

Robustness in NLP

Challenge:	Existing remediations have compromised accuracy or required full model re-training with each new class of attacks.
Approach:	They propose a method of retroactively adding resilience to misspellings to transformer-based NLP models and propose generating adversarial misspells using an approximate method.
Outcome:	The proposed method significantly reduces the cost needed to evaluate a model’s resilience to adversarial attacks.

Does Acceleration Cause Hidden Instability in Vision Language Models? Uncovering Instance-Level Divergence Through a Large-Scale Empirical Study (2025.emnlp-main)

Copied to clipboard

Yizheng Sun, Hao Li, Chang Xu, Hongpeng Zhou, Chenghua Lin, Riza Batista-Navarro, Jingyuan Sun

PokeMQA: Programmable knowledge editing for Multi-hop Question Answering (2024.acl-long)

Challenge:	Current acceleration evaluations focus on minimal overall performance degradation . however, accelerated models can exhibit significant changes in instance-level predictions .
Approach:	They investigate whether accelerated vision-Language Models can still give the same answers as before . they found that accelerated models changed original answers up to 20% of the time .
Outcome:	The results show that accelerated models changed their original answers up to 20% of the time.

Copied to clipboard

Hengrui Gu, Kaixiong Zhou, Xiaotian Han, Ninghao Liu, Ruobing Wang, Xin Wang

Simple Temporal Adaptation to Changing Label Sets: Hashtag Prediction via Dense KNN (2023.emnlp-main)

Challenge:	Multi-hop question answering (MQA) is one of the challenging tasks to evaluate machine’s comprehension and reasoning abilities, where large language models (LLMs) have widely achieved the human-comparable performance.
Approach:	They propose a framework to edit multi-hop question models to update model with up-to-date facts while avoiding expensive re-training or fine-tuning.
Outcome:	The proposed framework outperforms all competitors in multi-hop question answering tasks and consistently produces reliable reasoning process.

Copied to clipboard

Niloofar Mireshghallah, Nikolai Vogler, Junxian He, Omar Florez, Ahmed El-Kishky, Taylor Berg-Kirkpatrick

SHIELD: Defending Textual Neural Networks against Multiple Black-Box Adversarial Attacks with Stochastic Multi-Expert Patcher (2022.acl-long)

Challenge:	Existing methods to adapt to temporal change of user-generated social media data are stale without retraining.
Approach:	They propose a non-parametric dense retrieval technique to adapt to temporal change . they use a Twitter dataset to study temporal distribution shift in tweet-hashtag prediction .
Outcome:	The proposed method improves over the best static parametric baseline on a year-long Twitter dataset while avoiding costly re-training.

Copied to clipboard

Thai Le, Noseong Park, Dongwon Lee

Robustness in NLP

Challenge:	Existing methods to defend textual neural network models against adversarial attacks often require retraining and retrain . e.g., BERT, RoBERTa require great time and computation resources.
Approach:	They propose an algorithm that modifies and re-trains only the last layer of a textual NN and transforms it into a stochastic weighted ensemble of multi-expert prediction heads.
Outcome:	The proposed algorithm outperforms existing models against black-box attacks by 15%–70% . the proposed algorithm is based on a novel algorithm from software engineering .

Editing Common Sense in Transformers (2023.emnlp-main)

Copied to clipboard

Anshita Gupta, Debanjan Mondal, Akshay Sheshadri, Wenlong Zhao, Xiang Li, Sarah Wiegreffe, Niket Tandon

Editing Factual Knowledge in Language Models (2021.emnlp-main)

Challenge:	Currently, the performance of transformer-based model editing methods is limited to statements about encyclopedic knowledge with a single correct answer.
Approach:	They propose to improve MEMIT's model editing algorithm by varying edit tokens and improving the layer selection strategy to improve commonsense knowledge.
Outcome:	The MEMIT editing algorithm outperforms baseline models on PEP3k and 20Q datasets while fine-tuning baselines shows significant trade-offs.

Copied to clipboard

Nicola De Cao, Wilker Aziz, Ivan Titov

Few-shot Unified Question Answering: Tuning Models or Prompts? (2023.findings-emnlp)

Challenge:	KnowledgeEditor can be used to edit factual knowledge stored in Language Models without the need for expensive retraining or fine-tuning.
Approach:	They propose a method which edits factual knowledge implicitly stored in Language Models and uses it to fix 'bugs' and 'obvious errors' they train a hyper-network with constrained optimization to modify a fact without affecting the rest of the knowledge; the hyper-netzwork is then used to predict the weight update at test time.
Outcome:	The proposed method can be used to edit factual knowledge without retraining or fine-tuning and can fix 'bugs' or unexpected predictions without the need for expensive re-training or meta-learning.

Copied to clipboard

Srijan Bansal, Semih Yavuz, Bo Pang, Meghana Bhat, Yingbo Zhou

Outdated Issue Aware Decoding for Factual Knowledge Editing (2024.findings-acl)

Challenge:	Question-answering (QA) tasks investigate specific question types, knowledge domains, or reasoning skills, leading to specialized models catering to specific categories of QA tasks.
Approach:	They propose to use model and prompt tuning for unified QA in a low-resource setting to overcome drawbacks of unified models.
Outcome:	The proposed model and prompt tuning paradigms outperform model tuning in a few-shot setting with a good initialization and achieve a significant performance boost from pre-training in 'low-resource' setting.

Copied to clipboard

Zengkui Sun, Yijin Liu, Jiaan Wang, Fandong Meng, Jinan Xu, Yufeng Chen, Jie Zhou

Reasoning

Challenge:	Existing knowledge editing methods retain outdated responses for reasoning questions . naively retraining LLMs can be computationally intensive and can lead to catastrophic forgetting .
Approach:	They propose a simple yet effective decoding strategy to enhance edited models on reasoning questions.
Outcome:	The proposed method outDates ISsue aware deCOding (DISCO) to improve models on reasoning questions.

Controllable Semantic Parsing via Retrieval Augmentation (2021.emnlp-main)

Copied to clipboard

Panupong Pasupat, Yuan Zhang, Kelvin Guu

Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference? (2023.emnlp-main)

Challenge:	a mechanism for enacting behavior changes without expensive model re-training would be preferable.
Approach:	They propose a controllable semantic parser that retrieves related exemplars from a retrieval index and augments them to the query.
Outcome:	The proposed model can parse queries in a new domain, adapt predictions toward specified patterns, or adapt to new semantic schemas without re-training the model.

Copied to clipboard

Cheng Zhang, Jianyi Cheng, Ilia Shumailov, George Constantinides, Yiren Zhao

Challenge:	Existing quantisation methods mainly focus on 8-bit LLMs . a lack of scaling offsets in the quantisation process limits the use of LLM inference.
Approach:	They propose to use block quantisations to reduce scaling offsets in Large language models . they find that the block quantizations reduce scaling only from an arithmetic perspective .
Outcome:	The proposed methods reduce scaling offsets solely from an arithmetic perspective without additional treatments in the computational path.

You Only Need One Single Token to Refine Safety Alignment (2026.findings-acl)

Copied to clipboard

Wenqian Yu, Shuo Chen, Zhijiang Li, Zhipeng Wang, Jindong Gu

An LLM-based Framework for Biomedical Terminology Normalization in Social Media via Multi-Agent Collaboration (2025.coling-main)

Challenge:	Excessive safety can lead to over-refusal, where models reject harmful-looking yet benign queries, severely limiting utility.
Approach:	They propose a lightweight training-based approach that reshapes the distributions of harmful and benign samples within the model’s decision space by using a single-token prefix.
Outcome:	The proposed approach can distinguish between harmful and benign samples while keeping the model frozen.

Copied to clipboard

Yongqi Fan, Kui Xue, Zelin Li, Xiaofan Zhang, Tong Ruan

Reward-Augmented Decoding: Efficient Controlled Text Generation With a Unidirectional Reward Model (2023.emnlp-main)

Challenge:	Experimental results indicate that our approach exhibits competitive performance.
Approach:	They propose a tuning-free approach to normalize non-standard terms using large language models . they use a search engine and a domain knowledge base to expand the short texts into accurate descriptions .
Outcome:	The proposed approach is based on the "Recall and Re-rank" framework . it can be used to identify the standard term in a specified termbase for non-standardized mentions .

Copied to clipboard

Haikang Deng, Colin Raffel

SPIRIT: Patching Speech Language Models against Jailbreak Attacks (2025.emnlp-main)

Challenge:	Large language models generate text that is problematic or lacks a desired attribute.
Approach:	They propose a text generation procedure that uses a small unidirectional reward model to encourage a language model to generate text with certain properties.
Outcome:	The proposed procedure outperforms prior weighted decoding methods and matches state-of-the-art techniques that require additional training.

Copied to clipboard

Amirbek Djanibekov, Nurdaulet Mukhituly, Kentaro Inui, Hanan Aldarmaki, Nils Lukas

Robustness in NLP

Challenge:	Speech language models (SLMs) enable natural interactions via spoken instructions, which more effectively capture user intent by detecting nuances in speech.
Approach:	They propose post-hoc patching defenses to intervene during inference by modifying the SLM’s activations that improve robustness up to 99% with negligible impact on utility and without any re-training.
Outcome:	The proposed defenses improve robustness up to 99% with negligible impact on utility and (ii) without any re-training.

Harnessing Large Language Models as Post-hoc Correctors (2024.findings-acl)

Copied to clipboard

Zhiqiang Zhong, Kuangyu Zhou, Davide Mottin

Decouple knowledge from paramters for plug-and-play language modeling (2023.findings-acl)

Challenge:	Recent advances in Large Language Models (LLMs) have demonstrated their effectiveness in a wide range of tasks, including machine translation and commonsense reasoning.
Approach:	They propose a training-free framework that can work as a post-hoc corrector to propose corrections for ML models.
Outcome:	The proposed framework improves the performance of a number of models by up to 39% on text analysis and the challenging molecular predictions.

Copied to clipboard

Xin Cheng, Yankai Lin, Xiuying Chen, Dongyan Zhao, Rui Yan

Mind Your Neighbours: Leveraging Analogous Instances for Rhetorical Role Labeling for Legal Documents (2024.lrec-main)

Challenge:	Pre-trained language models (PLMs) have made impressive results in a wide range of NLP tasks.
Approach:	They propose a pre-training model with editable and scalable key-value memory and leverage knowledge in an explainable manner by knowledge retrieval in the pasted macro ‘MEMORY’.
Outcome:	The proposed model decouples the knowledge storage from model parameters with an editable and scalable key-value memory and leverages knowledge in an explainable manner by knowledge retrieval in the pasted macro ‘MEMORY’.

Copied to clipboard

Santosh T.y.s.s., Hassan Sarwat, Ahmed Mohamed Abdelaal Abdou, Matthias Grabmair

Challenge:	Rhetorical Role Labeling (RRL) of legal judgments presents challenges such as inferring sentence roles from context, interrelated roles, limited annotated data, and label imbalance.
Approach:	They propose techniques to enhance RRL performance by leveraging knowledge from semantically similar instances.
Outcome:	The proposed methods achieve remarkable improvements in challenging macro-F1 scores.

Beyond Fixed-Length Calibration for Post-Training Compression of LLMs (2025.findings-emnlp)

Copied to clipboard

Jaehoon Oh, Dokwan Oh

Challenge:	a recent study has demonstrated that the sequence length of calibration data plays a crucial role in the effectiveness of post-training compression methods.
Approach:	They propose a calibration technique that applies masking along the sequence axis to normalized hidden states.
Outcome:	The proposed method improves perplexity and zero-shot downstream tasks performance.

Mamba Drafters for Speculative Decoding (2025.findings-emnlp)

Copied to clipboard

Daewon Choi, Seunghyuk Oh, Saket Dingliwal, Jihoon Tack, Kyuyoung Kim, Woomin Song, Seojin Kim, Insu Han, Jinwoo Shin, Aram Galstyan, Shubham Katiyar, Sravan Babu Bodapati

BiasWipe: Mitigating Unintended Bias in Text Classifiers through Model Interpretability (2024.emnlp-main)

Challenge:	Existing drafters that use external drafters suffer from slower drafting while self-speculation methods use drafters tailored to the target model but require re-training.
Approach:	They propose a drafter based on a state space model, Mamba, as a solution that combines the best aspects of both approaches.
Outcome:	The proposed drafters outperform existing drafters while using less memory and maintaining their cross-model adaptability.

Copied to clipboard

Mamta Mamta, Rishikant Chigrupaatii, Asif Ekbal

Adapters Mixup: Mixing Parameter-Efficient Adapters to Enhance the Adversarial Robustness of Fine-tuned Pre-trained Text Classifiers (2024.emnlp-main)

Challenge:	Existing methods to mitigate unintended bias in social media platforms are re-training and adding extra parameters to the model.
Approach:	They propose a technique to mitigate unintended bias in language models by pruning the neuron weights responsible for univ bias.
Outcome:	The proposed technique achieves fairness by pruning the neuron weights responsible for unintended bias without loss of original performance.

Copied to clipboard

Tuc Nguyen, Thai Le

LADR: Locality-Aware Dynamic Rescue for Efficient Text-to-Image Generation with Diffusion Large Language Models (2026.acl-long)

Challenge:	Existing studies show that augmenting the training data of pre-trained language models with parametric fine-tuning methods can enhance their robustness under adversarial attacks.
Approach:	They propose an approach that fine-tunes PLMs with adapters and adversarial augmentation via mixup to leverage existing knowledge from a set of pre-known attacks.
Outcome:	The proposed approach achieves best trade-off between training efficiency and robustness under adversarial attacks compared to baselines on five downstream tasks across six varied black-box attacks and 2 PLMs.

Copied to clipboard

Chenglin Wang, Yucheng Zhou, Shuang Chen, Tao Wang, Kai Zhang

Visual Data in NLP

Challenge:	Existing methods for inference are expensive and lack spatial redundancy . Discrete Diffusion Language Models are a promising paradigm for multimodal generation .
Approach:	They propose a locality-aware dynamic rescue method that exploits spatial Markov property of images.
Outcome:	The proposed method achieves an approximate 4 speedup over baselines on four text-to-image generation benchmarks.

Let’s Fuse Step by Step: A Generative Fusion Decoding Algorithm with LLMs for Robust and Instruction-Aware ASR and OCR (2025.findings-acl)

Copied to clipboard

Chan-Jan Hsu, Yi-Chang Chen, Feng-Ting Liao, Pei-Chen Ho, Yu-Hsiang Wang, Po-Chun Hsu, Da-shan Shiu

Multimodality

Challenge:	Various fusion strategies have been explored for integration of large language models into multi-modal systems.
Approach:	They propose a framework for deep fusion decoding that integrates large language models into cross-modal text recognition systems.
Outcome:	The proposed framework surpasses cascaded methods in English and Mandarin, and significantly reduces WERs by 17.7%.

TiC-LM: A Web-Scale Benchmark for Time-Continual LLM Pretraining (2025.acl-long)

Copied to clipboard

Jeffrey Li, Mohammadreza Armandpour, Seyed Iman Mirzadeh, Sachin Mehta, Vaishaal Shankar, Raviteja Vemulapalli, Samy Bengio, Oncel Tuzel, Mehrdad Farajtabar, Hadi Pouransari, Fartash Faghri