Transactions of the Association for Computational Linguistics, Volume 14

Generative Induction of Dialogue Task Schemas with Streaming Refinement and Simulated Interactions (2026.tacl-1)

Challenge:	Existing methods for CR are encoder-only, decoder-based and asynchronous models.
Approach:	They propose a multilingual CR methodology which leverages decoder-only LLMs to handle overt and zero mentions.
Outcome:	The proposed model outperforms the leading multilingual CR model by 2 percentage points across all languages in the CorefUD v1.2 dataset.

Copied to clipboard

James D. Finch, Yasasvi Josyula, Jinho D. Choi

Dialogue Systems & Conversational Agents

Challenge:	Slot Schema Induction (SSI) is a task-oriented dialogue (TOD) system that allows for automatic identification of information slots from unlabeled data.
Approach:	They propose a language model that incrementally constructs and refines a slot schema over a stream of dialogue data and then automatically creates high-quality state labels.
Outcome:	The proposed method creates high-quality state labels for novel task domains and improves evaluation metrics.

Localizing Factual Inconsistencies in Attributable Text Generation (2026.tacl-1)

Copied to clipboard

Arie Cattan, Paul Roit, Shiyue Zhang, David Wan, Roee Aharoni, Idan Szpektor, Mohit Bansal, Ido Dagan

Text Generation

Challenge:	Existing methods for detecting hallucinations in model-generated texts fail to pinpoint errors.
Approach:	They propose a formalism for localizing factual inconsistencies in attributable text generation . they propose to decompose the generated text into simple question-answer pairs .
Outcome:	The proposed method achieves substantial inter-annotator agreement while achieving a substantial consistency score.

What Can String Probability Tell Us About Grammaticality? (2026.tacl-1)

Copied to clipboard

Jennifer Hu, Ethan Gotlieb Wilcox, Siyuan Song, Kyle Mahowald, Roger P. Levy

Simulating Hard Attention Using Soft Attention (2026.tacl-1)

Challenge:	linguistic theories have argued that language models have largely achieved grammatical competence, but they will assign non-zero probability to all strings.
Approach:	They propose a theoretical framework for analyzing string probabilities in linguistics based on simple assumptions about the generative process of corpus data.
Outcome:	The proposed framework makes three predictions using 280K sentence pairs in English and Chinese.

Copied to clipboard

Andy Yang, Lena Strobl, David Chiang, Dana Angluin

On the Limitations of Language-targeted Pruning: Investigating the Calibration Language Impact in Multilingual LLM Pruning (2026.tacl-1)

Challenge:	a central element of hard attention is attention, which computes a weighted average of values from all unmasked positions.
Approach:	They propose transformers that can simulate hard attention by using temperature scaling and positional embeddings.
Outcome:	The proposed transformers can effectively focus all attention on a subset of positions.

Copied to clipboard

Simon Kurz, Jian-Jia Chen, Lucie Flek, Zhixue Zhao

MultiBLiMP 1.0: A Massively Multilingual Benchmark of Linguistic Minimal Pairs (2026.tacl-1)

Challenge:	Recent advances in large language model pruning have shown high predictive performance in post-training settings.
Approach:	They conduct an empirical study on the performance and internal representation changes associated with pruning multilingual models for monolingual applications.
Outcome:	The proposed pruning methods retain perplexity and yield high signal-to-noise ratios, but not consistently improve downstream tasks.

Copied to clipboard

Jaap Jumelet, Leonie Weissweiler, Joakim Nivre, Arianna Bisazza

PiKGL: Leveraging Pruned Knowledge Graphs for Explainable Stance Detection (2026.tacl-1)

Challenge:	MultiBLiMP 1.0 is a massively multilingual benchmark of linguistic minimal pairs covering 101 languages and 2 types of subject-verb agreement.
Approach:	They propose to use multilingual benchmarks to evaluate linguistic minimal pairs in 101 languages and 2 types of subject-verb agreement to create the minimal pairs.
Outcome:	The proposed benchmark covers 101 languages and 2 types of subject-verb agreement, and contains more than 128,000 minimal pairs.

Copied to clipboard

Bingbing Wang, Jingjie Lin, Zhixin Bai, Xintong Song, Qianlong Wang, Min Yang, Xi Zeng, Jing Li, Ruifeng Xu

Sentiment Analysis

Challenge:	Experimental results demonstrate that a Pruned interpretable knowledge Graph Learning framework for explainable stance detection is state-of-the-art for social media stance prediction.
Approach:	They propose a Pruned interpretable knowledge Graph Learning framework for explainable stance detection that incorporates commonsense knowledge and prunes redundant information to ensure precision and minimize noise.
Outcome:	The proposed framework achieves state-of-the-art on three public datasets.

Do Large Multimodal Models Solve Caption Generation for Scientific Figures? Lessons Learned from SciCap Challenge 2023 (2026.tacl-1)

Copied to clipboard

Ting-Yao Hsu, Yi-Li Hsu, Shaurya Rohatgi, Chieh-Yang Huang, Ho Yin Sam Ng, Ryan Rossi, Sungchul Kim, Tong Yu, Lun-Wei Ku, Clyde Lee Giles, Ting-Hao Huang

Multimodality

Challenge:	SciCap dataset launched in 2021 aims to generate high-quality captions for scientific figures.
Approach:	They propose to use the SciCap dataset to develop models for captioning diverse figure types across various academic fields.
Outcome:	The proposed models showed impressive performance on the SciCap dataset and in various vision-and-language tasks.

Accelerating Language Model Workflows with Prompt Choreography (2026.tacl-1)

Copied to clipboard

TJ Bai, Jason Eisner

Aligned Probing: Relating Toxic Behavior and Model Internals (2026.tacl-1)

Challenge:	Large language models are increasingly deployed in multi-agent workflows that require multiple agents to encode the same prompt from scratch.
Approach:	They propose a framework that maintains a dynamic, global KV cache that allows agents to attend to arbitrary, reordered subsets of previously encoded messages.
Outcome:	The proposed framework significantly reduces per-message latency (2.0–6.2 faster time-to-first-token) and achieves substantial speedups (>2.2) in some workflows dominated by redundant computation.

Copied to clipboard

Andreas Waldis, Vagrant Gautam, Anne Lauscher, Dietrich Klakow, Iryna Gurevych

Self-Consistency Falls Short! The Adverse Effects of Positional Bias on Long-Context Problems (2026.tacl-1)

Challenge:	Language models (LMs) may produce toxic text that contains hate speech, insults, or vulgarity, even when prompted with innocuous text.
Approach:	They propose an interpretability framework that aligns the behavior of language models based on their outputs and internal representations.
Outcome:	The proposed framework bridges behavioral and internal perspectives for toxicity for the first time.

Copied to clipboard

Adam Byerly, Daniel Khashabi

Challenge:	Existing approaches to improve long-context understanding of large language models lack scalability and stability.
Approach:	They challenge the assumption that SC’s benefits generalize to long-context settings . they find that persistent position bias degrades performance on long-consistency tasks .
Outcome:	The proposed approach fails to improve and actively degrades performance on long-context tasks.

IssueBench: Millions of Realistic Prompts for Measuring Issue Bias in LLM Writing Assistance (2026.tacl-1)

Copied to clipboard

Paul Röttger, Musashi Hinck, Valentin Hofmann, Kobi Hackenburg, Valentina Pyatkin, Faeze Brahman, Dirk Hovy

Fine-tuning Large Language Models with Limited Data: A Survey and Practical Guide (2026.tacl-1)

Challenge:	Large language models are helping millions of users write texts about diverse issues . issue bias is where an LLM tends to present just one perspective on a given issue .
Approach:	They construct a set of 2.49m realistic English-language prompts to measure issue bias in LLM writing assistance using 3.9k templates and 212 political issues from real user interactions.
Outcome:	The proposed model aligns more with US Democrat than Republican voter opinion on a subset of issues.

Copied to clipboard

Marton Szep, Daniel Rueckert, Rüdiger von Eisenhart-Rothe, Florian Hinterwimmer

VoiceBench: Benchmarking LLM-Based Voice Assistants (2026.tacl-1)

Challenge:	Pre-trained language models provide strong foundations, but effective adaptation under data scarcity requires efficient and efficient fine-tuning techniques.
Approach:	They propose to review parameter-efficient fine-tuning techniques that lower training and deployment costs and domain and cross-lingual adaptation methods for both encoder and decoder models.
Outcome:	The proposed techniques lower training and deployment costs, domain and cross-lingual adaptation methods, and model specialization strategies.

Copied to clipboard

Yiming Chen, Xianghu Yue, Chen Zhang, Xiaoxue Gao, Robby T. Tan, Haizhou Li

Speech & Audio in NLP

Challenge:	Recent advances in large language models (LLMs) have enabled real-time speech interactions through LLMs.
Approach:	They propose a benchmark specifically designed to assess LLM-based voice assistants.
Outcome:	The proposed benchmark measures the performance of LLM-based voice assistants across eight tasks.

MERLIN: A Testbed for Multilingual Multimodal Entity Recognition and Linking (2026.tacl-1)

Copied to clipboard

Sathyanarayanan Ramamoorthy, Vishwa Shah, Simran Khanuja, Zaid Sheikh, Shan Jie, Ann Chia, Shearman Chua, Graham Neubig

Information Extraction & Text Mining

Challenge:	Existing methods for multilingual entity linking are limited by textual contexts and limited resources.
Approach:	They propose a testbed system for multilingual multimodal entity linking using BBC news articles paired with corresponding images in five languages.
Outcome:	The proposed system improves accuracy for entities with ambiguous textual contexts and models with weak multilingual abilities.

🧑‍🍳 Cooking Up Creativity: Enhancing LLM Creativity through Structured Recombination (2026.tacl-1)

Copied to clipboard

Moran Mizrahi, Chen Shani, Gabriel Stanovsky, Dan Jurafsky, Dafna Shahaf

Challenge:	Large Language Models excel at many tasks, yet struggle to generate truly creative ideas.
Approach:	They propose a novel approach that enhances Large Language Models' creativity by manipulating structured representations of existing ideas.
Outcome:	The proposed model outperforms GPT-4o in novelty and diversity and outperformed GPT-0 in creative generation.

Generating Visual Stories with Grounded and Coreferent Characters (2026.tacl-1)

Copied to clipboard

Danyang Liu, Mirella Lapata, Frank Keller

Visual Data in NLP

Challenge:	Experimental results show that our model generates visual stories with consistent and coreferent character mentions compared to baselines and state-of-the-art systems.
Approach:	They propose a character-centric approach to visual story generation that uses visual and textual character coreference chains to enrich the VIST benchmark.
Outcome:	The proposed model generates visual stories with consistent and coreferent character mentions compared to baselines and state-of-the-art systems.

Automatic Reviewers Fail to Detect Faulty Reasoning in Research Papers: A New Counterfactual Evaluation Framework (2026.tacl-1)

Copied to clipboard

Nils Dycke, Iryna Gurevych

Reasoning

Challenge:	Large Language Models (LLMs) are increasingly used as fully automatic review generators (ARGs).
Approach:	They propose a fully automated counterfactual evaluation framework that isolates and tests a core review skill that underpins high-quality peer review: detecting faulty research logic.
Outcome:	The proposed framework isolates and tests a range of ARG approaches and shows that flaws in research logic have no significant effect on their output reviews.

Can LLMs Automate Fact-Checking Article Writing? (2026.tacl-1)

Copied to clipboard

Dhruv Sahnan, David Corney, Irene Larraz, Giovanni Zagni, Ruben Miguez, Zhuohan Xie, Iryna Gurevych, Elizabeth Churchill, Tanmoy Chakraborty, Preslav Nakov

PsyMem: Fine-grained Psychological Alignment and Explicit Memory Control for Advanced Role-Playing LLMs (2026.tacl-1)

Challenge:	Existing tools for automatic fact-checking produce little or no justification for their assessments . 80% of american adults on major social media platforms regularly encounter news-related content .
Approach:	They propose to extend automatic fact-checking pipeline with automatic generation of full fact- checking articles.
Outcome:	The proposed framework outperforms existing frameworks but lags behind expert-written articles.

Copied to clipboard

Xilong Cheng, Yunxiao Qin, Yuting Tan, Zhengnan Li, Ye Wang, Hongjiang Xiao, Yuan Zhang

Challenge:	Existing role-playing models rely on superficial textual descriptions or simplistic metrics, inadequately modeling both intrinsic and extrinsic character dimensions.
Approach:	They propose a framework that integrates fine-grained psychological attributes and explicit memory control for role-playing.
Outcome:	The proposed framework outperforms baseline models in human-likeness and character fidelity.

Ev2R: Evaluating Evidence Retrieval in Automated Fact-Checking (2026.tacl-1)

Copied to clipboard

Mubashara Akhtar, Michael Schlichtkrull, Andreas Vlachos

A Context-aware Framework for Translation-mediated Conversations (2026.tacl-1)

Challenge:	Current methods for automated fact-checking rely on relying on other evaluation metrics and closed knowledge sources.
Approach:	They propose a method which combines evidence evaluation with verdict-level proxy scoring.
Outcome:	The proposed method outperforms existing methods in accuracy and robustness against human ratings and adversarial tests.

Copied to clipboard

José Pombal, Sweta Agrawal, Emmanouil Zaranis, Patrick Fernandes, André F. T. Martins

Can Language Models Learn Typologically Implausible Languages? (2026.tacl-1)

Challenge:	Existing systems that bridge language barriers can introduce errors leading to misunderstandings and conversation breakdown.
Approach:	They propose a framework to integrate contextual information into automatic translation systems . they validate the framework on customer chat and user-assistant interaction .
Outcome:	The proposed framework consistently produces better translations than state-of-the-art systems on two task-oriented domains.

Copied to clipboard

Tianyang Xu, Tatsuki Kuribayashi, Yohei Oseki, Ryan Cotterell, Alex Warstadt

Can Large Language Models Generalize Analogy Solving Like Children Can? (2026.tacl-1)

Challenge:	Language models provide a naturalistic framework for studying artificial language learning . authors: typological universals and tendencies are thought to be caused by a learning bias .
Approach:	They propose to train LMs on highly naturalistic counterfactual versions of English and Japanese . they show that LM learn subtly implausible languages more slowly .
Outcome:	The proposed language models learn subtly implausible languages more slowly compared to human models . the findings suggest that LMs exhibit typologically aligned learning preferences .

Copied to clipboard

Claire E. Stevenson, Alexandra Pafford, Han L. J. van der Maas, Melanie Mitchell

Dissecting GraphRAG: A Modular Analysis of Knowledge Structuring for Factoid Question Answering (2026.tacl-1)

Challenge:	Recent research shows that large language models can solve various forms of analogies.
Approach:	They investigated whether large language models can generalize analogy solving to other domains . they found that LLMs struggle with robust human-like analogical transfer .
Outcome:	The results show that large language models can solve analogies in the Latin alphabet, Greek alphabet, and far transfer domains.

Copied to clipboard

Noriki Nishida, Rumana Ferdous Munne, Shanshan Liu, Narumi Tokunaga, Yuki Yamagata, Fei Cheng, Kouji Kozaki, Yuji Matsumoto

Information Extraction & Text Mining

Challenge:	GraphRAG integrates structured knowledge graphs into question answering . high-quality triple extraction is critical, but lacks granularity and topical coherence . large language models suffer from inherent limitations in their internalized knowledge .
Approach:	They evaluate module-level design choices in GraphRAG for retrieval-augmented generation . they find that triple extraction is critical for accurate and comprehensive retrieval .
Outcome:	The proposed framework outperforms other retrieval-augmented generation frameworks in accuracy and efficiency.

Cross-layer Attention Sharing for Pre-trained Large Language Models (2026.tacl-1)

Copied to clipboard

Yongyu Mu, Yuzhang Wu, Yuchun Fan, Chenglong Wang, Hengyu Li, Jiali Zeng, Qiaozhi He, Murun Yang, Fandong Meng, Jie Zhou, Tong Xiao, Jingbo Zhu

Beyond One-Size-Fits-All: Inversion Learning for Highly Effective NLG Evaluation Prompts (2026.tacl-1)

Challenge:	Existing studies focus on compressing the Key-Value cache or grouping attention heads, while overlooking redundancy between layers.
Approach:	They propose a lightweight substitute for self-attention in well-trained LLMs that uses feed-forward networks to align attention heads between adjacent layers and low-rank matrices to approximate differences in layer-wise attention weights.
Outcome:	The proposed model reduces redundancy by sharing weights across layers while maintaining high response quality while reducing redundant calculations within 53% 84% of the total layers.

Copied to clipboard

Hanhua Hong, Chenghao Xiao, Yang Wang, Yiqi Liu, Wenge Rong, Chenghua Lin