Papers by Tianyi Zhou

Can LLMs Speak For Diverse People? Tuning LLMs via Debate to Generate Controllable Controversial Statements (2024.findings-acl)

Challenge:	Recent advances in natural language tasks leverage the emergent In-Context Learning ability of pretrained Large Language Models (LLMs).
Approach:	They propose a framework for exemplar selection for in-context learning that uses a pool-based active learning approach to select Diverse and informative exemplars from the target tasks’ unlabeled pool.
Outcome:	The proposed framework outperforms existing methods for data annotation and similarity-based methods for test query-specific exemplar retrieval on 7 different NLP datasets and 5 LLMs of varying complexities.

Copied to clipboard

Ming Li, Jiuhai Chen, Lichang Chen, Tianyi Zhou

Challenge:	Existing LLMs lack sufficient controllability to generate statements supporting diverse or even controversial perspectives.
Approach:	They develop a pipeline that fine tunes LLMs to generate statements generated via debate.
Outcome:	The proposed pipeline improves the controllability of LLMs in generating statements supporting an argument the user defined in the prompt.

ChatMap: Mining Human Thought Processes for Customer Service Chatbots via Multi-Agent Collaboration (2025.findings-acl)

Copied to clipboard

Xinyi Jiang, Tianyi Hu, Yuheng Qin, Guoming Wang, Zhou Huan, Kehan Chen, Gang Huang, Rongxing Lu, Siliang Tang

Dialogue Systems & Conversational Agents

Challenge:	Existing methods for enhancing dialogue performance rely on summarizing behavior . e-commerce chatbots need to align their dialogue strategies with human behavior to achieve coherent, human-like conversations with customers.
Approach:	They propose a method to extract core patterns from dialogue data and integrate them into models by mining service thought processes using a multi-agent aPproach.
Outcome:	The proposed method outperforms manual methods and outperfies baselines on Taobao in China.

Schoenfeld’s Anatomy of Mathematical Reasoning by Language Models (2026.acl-long)

Copied to clipboard

Ming Li, Chenrui Fan, Yize Cheng, Soheil Feizi, Tianyi Zhou

Reasoning

Challenge:	Large language models expose reasoning traces, yet their underlying cognitive structure and steps remain difficult to identify and analyze beyond surface-level statistics.
Approach:	They propose a framework that explicitly abstracts reasoning traces into functional reasoning steps such as Analysis, Explore, Implement, Verify, etc.
Outcome:	The proposed framework reveals reproducible thinking dynamics and structural differences between reasoning and non-reasoning models, which are not apparent from token-level views.

From Lists to Emojis: How Format Bias Affects Model Alignment (2025.acl-long)

Copied to clipboard

Xuanchang Zhang, Wei Xiong, Lichang Chen, Tianyi Zhou, Heng Huang, Tong Zhang

Multimodality

Challenge:	Format biases in reinforcement learning from human feedback are underexplored . despite its effectiveness, RLHF faces challenges, including policy and regulatory constraints .
Approach:	They extend the study of preference biases beyond verbosity bias to a wider range of format biase . they show that with a small amount of biased data, they can inject significant bias into the reward model .
Outcome:	The proposed approach can be easily exploited by large language models to achieve higher rankings on popular benchmarks like AlpacaEval and LMSYS Chatbot Arena.

Sparser Mixture-of-Adapters with Cross-Layer Generalization (2025.naacl-long)

Copied to clipboard

Ziyue Li, Tianyi Zhou

Green & Sustainable NLP

Challenge:	Existing methods for training large language models do not allow sharing adapters across layers . existing methods do not support sharing adapter pools, leading to redundancy and poor generalization .
Approach:	They propose a mixture-of-adapter framework that trains a pool of lightweight adapters at each layer and selects the most suitable ones for each input.
Outcome:	The proposed framework reduces active adapters by over 85% while boosting task accuracy.

Multiple LLM Agents Debate for Equitable Cultural Alignment (2025.acl-long)

Copied to clipboard

Dayeon Ki, Rachel Rudinger, Tianyi Zhou, Marine Carpuat

Challenge:	Recent efforts focus on single-LLM, single-turn generation approaches, but it can be challenging for any single model to support all cultures equally well.
Approach:	They propose to exploit the complementary strengths of multiple LLMs to promote cultural adaptability.
Outcome:	The proposed model improves accuracy and cultural group parity over single-LLM models.

1+1>2: Can Large Language Models Serve as Cross-Lingual Knowledge Aggregators? (2024.emnlp-main)

Copied to clipboard

Yue Huang, Chenrui Fan, Yuan Li, Siyuan Wu, Tianyi Zhou, Xiangliang Zhang, Lichao Sun

CogKTR: A Knowledge-Enhanced Text Representation Toolkit for Natural Language Understanding (2022.emnlp-demos)

Challenge:	Large Language Models (LLMs) have been recognized for their impressive capabilities in natural language processing (NLP).
Approach:	They propose a method to enhance the multilingual performance of Large Language Models by aggregating knowledge from diverse languages.
Outcome:	The proposed method reduces the performance disparity across languages and offers valuable insights for further exploration.

Copied to clipboard

Zhuoran Jin, Tianyi Men, Hongbang Yuan, Yuyang Zhou, Pengfei Cao, Yubo Chen, Zhipeng Xue, Kang Liu, Jun Zhao

Token Dropping for Efficient BERT Pretraining (2022.acl-long)

Challenge:	Existing knowledge-enhanced methods are limited to knowledge-intensive tasks.
Approach:	They propose a knowledge-enhanced text representation toolkit for natural language understanding . it combines knowledge acquisition, knowledge representation, knowledge injection and knowledge application .
Outcome:	The proposed toolkit supports knowledge acquisition, knowledge representation, knowledge injection, and knowledge application.

Copied to clipboard

Le Hou, Richard Yuanzhe Pang, Tianyi Zhou, Yuexin Wu, Xinying Song, Xiaodan Song, Denny Zhou

Mitigating Inconsistencies in Multimodal Sentiment Analysis under Uncertain Missing Modalities (2022.emnlp-main)

Challenge:	Existing methods to accelerate pretraining of transformer-based models are computationally expensive and degrade performance on downstream tasks.
Approach:	They propose a "token dropping" method to accelerate the pretraining of transformer-based models by 25% . they leverage the already built-in masked language modeling loss to identify unimportant tokens with practically no computational overhead.
Outcome:	The proposed method reduces the pretraining cost of BERT models by 25% while achieving similar overall performance on downstream tasks.

Copied to clipboard

Jiandian Zeng, Jiantao Zhou, Tianyi Liu

AutoHallusion: Automatic Generation of Hallucination Benchmarks for Vision-Language Models (2024.findings-emnlp)

Challenge:	Existing studies ignore the inconsistency phenomenon of missing modality in multimodal sentiment analysis . neglect of missing modalities may lead to incorrect semantic results .
Approach:	They propose an ensemble-based Missing Modality Reconstruction network to detect and recover missing modality features.
Outcome:	The proposed method is superior to existing methods on CMU-MOSI and IEMOCAP datasets.

Copied to clipboard

Xiyang Wu, Tianrui Guan, Dianqi Li, Shuaiyi Huang, Xiaoyu Liu, Xijun Wang, Ruiqi Xian, Abhinav Shrivastava, Furong Huang, Jordan Boyd-Graber, Tianyi Zhou, Dinesh Manocha

Reasoning

Challenge:	Large vision-language models are prone to hallucinations, where contextual cues in an image can trigger the language module to produce overconfident and incorrect reasoning about abnormal or hypothetical objects.
Approach:	They propose to automate the generation of hallucination-related questions using images . they propose to use three image manipulation strategies to induce hallucinosity .
Outcome:	The proposed approach reduces human bias in crafting such examples and improves accuracy.

Value Residual Learning (2025.acl-long)

Copied to clipboard

Zhanchao Zhou, Tianyi Wu, Zhiyun Jiang, Fares Obeid, Zhenzhong Lan

TASA: Deceiving Question Answering Models by Twin Answer Sentences Attack (2022.emnlp-main)

Challenge:	Existing decoder-only transformers fail to preserve initial token-level information in deeper layers.
Approach:	They propose a new architecture that incorporates value residual connections in addition to hidden state residuals.
Outcome:	The proposed architecture reduces KV cache size by nearly half with only a small performance penalty and can be integrated with other KV-efficient methods.

Copied to clipboard

Yu Cao, Dianqi Li, Meng Fang, Tianyi Zhou, Jun Gao, Yibing Zhan, Dacheng Tao

Robustness in NLP

Challenge:	Existing adversarial models rely on keyword matching and ignore relevant contextual relations for answer prediction.
Approach:	They propose to use keyword matching to attack model with two biases that rely on a perturbed answer sentence and a distracting answer sentence to misguide model.
Outcome:	The proposed method produces fluent and grammatical adversarial contexts while maintaining gold answers.

Human-in-the-loop Schema Induction (2023.acl-demo)

Copied to clipboard

Tianyi Zhang, Isaac Tham, Zhaoyi Hou, Jiaxuan Ren, Leon Zhou, Hainiu Xu, Li Zhang, Lara Martin, Rotem Dror, Sha Li, Heng Ji, Martha Palmer, Susan Windisch Brown, Reece Suchocki, Chris Callison-Burch

Structured Data in NLP

Challenge:	Existing approaches to event-centric natural language understanding (NLU) have been limited to linear and temporal ones.
Approach:	They propose a human-in-the-loop schema induction system powered by GPT-3 . they show that it transfers to new domains more easily than previous approaches .
Outcome:	The proposed system transfers to new domains more easily than previous approaches and reduces human curation.

RMTBench: Benchmarking LLMs Through Multi-Turn User-Centric Role-Playing (2025.findings-emnlp)

Copied to clipboard

Hao Xiang, Tianyi Tang, Yang Su, Bowen Yu, An Yang, Fei Huang, Yichang Zhang, Yaojie Lu, Hongyu Lin, Xianpei Han, Jingren Zhou, Junyang Lin, Le Sun

Dialogue Systems & Conversational Agents

Challenge:	Existing benchmarks focus on character-centric approach and fail to reflect real-world applications.
Approach:	RMTBench is a user-centric bilingual role-playing benchmark featuring 80 diverse characters and over 8,000 dialogue rounds.
Outcome:	RMTBench features 80 diverse characters and over 8,000 dialogue rounds.

Merging Experts into One: Improving Computational Efficiency of Mixture of Experts (2023.emnlp-main)

Copied to clipboard

Shwai He, Run-Ze Fan, Liang Ding, Li Shen, Tianyi Zhou, Dacheng Tao

Green & Sustainable NLP

Challenge:	Extensive experiments show that MEO significantly improves computational efficiency . compared to dense networks, sparsely activated networks only employ a few parameters for each input .
Approach:	They propose a method that merges multiple experts into one to reduce computation costs . they demonstrate that a sparse Mixture of Experts (MoE) can reduce the cost by activating a small subset of parameters for each input .
Outcome:	The proposed approach reduces the computational cost to that of a single expert by 83.3% compared to 82.6% in vanilla MoE.

How Many Demonstrations Do You Need for In-context Learning? (2023.findings-emnlp)

Copied to clipboard

Jiuhai Chen, Lichang Chen, Chen Zhu, Tianyi Zhou

RuleR: Improving LLM Controllability by Rule-based Data Recycling (2025.naacl-short)

Challenge:	Large language models (LLMs) are capable of complex reasoning when given a few input-output demos.
Approach:	They use fewer input-output demos for each test query to study ICL . they do not observe significant degradation when using only one randomly chosen demo .
Outcome:	The proposed model outperforms multi-demo models on the tasks in 2022.

Copied to clipboard

Ming Li, Han Chen, Chenguang Wang, Dang Nguyen, Dianqi Li, Tianyi Zhou

Challenge:	Existing supervised fine-tuning datasets are composed of general instructions without userspecified constraints.
Approach:	They propose a data augmentation method incorporating multiple constraints into the original data samples according to predefined rules to create new training tasks.
Outcome:	The proposed method improves LLM controllability while maintaining general instruction-following capabilities.

Regularized Attentive Capsule Network for Overlapped Relation Extraction (2020.coling-main)

Copied to clipboard

Tianyi Liu, Xiangyu Lin, Weijia Jia, Mingliang Zhou, Wei Zhao

Information Extraction & Text Mining

Challenge:	Existing methods to extract relations from distant supervision contain low-quality instances with noisy words and overlapped relations.
Approach:	They propose a Regularized Attentive Capsule Network to better identify overlapped relations in informal sentences . they embed multi-head attention into the capsule network as the low-level capsules .
Outcome:	Extensive experiments show that the proposed model improves relation extraction.

From Quantity to Quality: Boosting LLM Performance with Self-Guided Data Selection for Instruction Tuning (2024.naacl-long)

Copied to clipboard

Ming Li, Yong Zhang, Zhitao Li, Jiuhai Chen, Lichang Chen, Ning Cheng, Jianzong Wang, Tianyi Zhou, Jing Xiao

Selective Reflection-Tuning: Student-Selected Data Recycling for LLM Instruction-Tuning (2024.findings-acl)

Challenge:	Large Language Models (LLMs) have revolutionized the landscape of artificial intelligence.
Approach:	They propose a self-guided method to identify and select cherry samples from open-source datasets, minimizing manual curation and potential cost for instruction tuning an LLM.
Outcome:	The proposed method enables LLMs to identify discrepancies between expected responses and intrinsic generation capability, and a marked uptick in model training efficiency.

Copied to clipboard

Ming Li, Lichang Chen, Jiuhai Chen, Shwai He, Jiuxiang Gu, Tianyi Zhou

How Instruction and Reasoning Data shape Post-Training: Data Quality through the Lens of Layer-wise Gradients (2026.acl-long)

Challenge:	Instruction tuning is critical to large language models but its success heavily relies on the training data quality.
Approach:	They propose a paradigm that synergizes a teacher LLM’s reflection and introspection with the data selection capability of the student LLM to automatically refine existing instruction-tuning data.
Outcome:	The proposed method achieves much stronger and top-tier 7B and 13B LLMs without collecting brand-new data.

Copied to clipboard

Ming Li, Yanhong Li, Ziyue Li, Tianyi Zhou

Mosaic-IT: Cost-Free Compositional Data Synthesis for Instruction Tuning (2025.findings-acl)

Challenge:	Spectral properties of low/high-quality instruction and reasoning data are used to explain finetuning dynamics in large language models.
Approach:	They propose to analyze layer-wise gradients induced by low/high-quality instruction and reasoning data for LLM post-training.
Outcome:	The results show that higher-quality data are associated with lower nuclear norms and higher effective ranks.

Copied to clipboard

Ming Li, Pei Chen, Chenguang Wang, Hongyu Zhao, Yijun Liang, YuPeng Hou, Fuxiao Liu, Tianyi Zhou

Do great minds think alike? Investigating Human-AI Complementarity in Question Answering with CAIMIRA (2024.emnlp-main)

Challenge:	Current instruction tuning relies on teacher models or human intervention to generate and refine the instructions and responses for training, which are costly, non-sustainable, and may lack diversity.
Approach:	They propose a human/model-free compositional data synthesis method that can create rich and diverse augmentations from existing instruction tuning data to enhance large language models.
Outcome:	The proposed method improves performance over benchmarks and reduces training costs by 80% compared with original instruction tuning.

Copied to clipboard

Maharshi Gor, Hal Daumé Iii, Tianyi Zhou, Jordan Boyd-Graber

Question Answering

Challenge:	Recent advances in large language models have led to claims of AI surpassing humans in QA tasks . authors: models are purportedly acing tests that many humans find challenging .
Approach:	They propose a framework that enables quantitative assessment and comparison of problem-solving abilities in QA agents.
Outcome:	The proposed framework uncovers distinctficiency patterns in knowledge domains and reasoning skills.

Skill Discovery for Software Scripting Automation via Offline Simulations with LLMs (2026.findings-eacl)

Copied to clipboard

Paiheng Xu, Gang Wu, Xiang Chen, Tong Yu, Chang Xiao, Franck Dernoncourt, Tianyi Zhou, Wei Ai, Viswanathan Swaminathan

Multimodality

Challenge:	Large Language Models (LLMs) can generate code from natural language queries, but runtime code generation is limited due to unverified code, security risks, longer response times, and higher computational costs.
Approach:	They propose an offline simulation framework to curate a software-specific skillset by exploiting large language models and publicly available scripting guides.
Outcome:	The proposed framework significantly improves automation success rates, reduces response time, and saves runtime token costs compared to traditional runtime code generation.

DrAttack: Prompt Decomposition and Reconstruction Makes Powerful LLMs Jailbreakers (2024.findings-emnlp)

Copied to clipboard

Xirui Li, Ruochen Wang, Minhao Cheng, Tianyi Zhou, Cho-Jui Hsieh

Meta-Task Prompting Elicits Embeddings from Large Language Models (2024.acl-long)

Challenge:	Existing jailbreaking methods view a malicious prompt as a whole but they are not effective at reducing LLMs’ attention on combinations of words with malice.
Approach:	They propose an automatic prompt Decomposition and Reconstruction framework for jailbreaking Attack that decomposes a malicious prompt into separate sub-prompts and reassembles them implicitly by In-Context Learning.
Outcome:	The proposed framework reduces LLMs' attention on malice words by presenting them to LLM in a fragmented form, addressing these limitations and improving attack effectiveness.

Copied to clipboard

Yibin Lei, Di Wu, Tianyi Zhou, Tao Shen, Yu Cao, Chongyang Tao, Andrew Yates

ATLAS: Agent Tuning via Learning Critical Steps (2025.findings-acl)

Challenge:	Existing methods for large language modeling are based on task-related instructions or prompts.
Approach:	They propose a method for generating high-quality sentence embeddings from Large Language Models (LLMs) using meta-task prompts.
Outcome:	The proposed method produces high-quality sentences without fine-tuning . it excels on STS benchmarks and in downstream tasks, surpassing models with similar prompts .

Copied to clipboard

Zhixun Chen, Ming Li, Yuxuan Huang, Yali Du, Meng Fang, Tianyi Zhou

SpecHub: Provable Acceleration to Multi-Draft Speculative Decoding (2024.emnlp-main)

Challenge:	Existing agent tuning approaches employ supervised finetuning on entire expert trajectories, but behavior-cloning of full traitories introduces expert bias and weakens generalization to states not covered by the expert data.
Approach:	They propose a method that finetunes LLMs on critical steps in expert trajectories and identifies and finetuns them on these steps with reduced costs.
Outcome:	The proposed method outperforms existing methods and open-source LLM agents on only 30% critical steps in extensive experiments.

Copied to clipboard

Ryan Sun, Tianyi Zhou, Xun Chen, Lichao Sun

Neural Relation Extraction via Inner-Sentence Noise Reduction and Transfer Learning (D18-1)

Challenge:	Large Language Models (LLMs) have limited inference speed due to sequential token generation . Spechub is a novel, efficient sampling-verification method for MDSD that improves acceptance rates with only linear computational overhead.
Approach:	They propose a method that uses a smaller draft model to generate multiple token sequences . Spechub generates 0.05-0.27 and 0.02-0.16 more tokens per step than RRS and RRS without replacement .
Outcome:	The proposed method improves acceptance rates with only linear computational overhead.

Copied to clipboard

Tianyi Liu, Xinsong Zhang, Wanhao Zhou, Weijia Jia

Improving Long-Tail Relation Extraction with Collaborating Relation-Augmented Attention (2020.coling-main)

Challenge:	Existing methods for extracting relations are slow and lack precision . a novel approach to extract relations is proposed to reduce noise between sentences .
Approach:	They propose a word-level distant supervised approach for relation extraction using New York Times and Freebase.
Outcome:	The proposed method improves the area of precision/call(PR) from 0.35 to 0.39 over the state-of-the-art methods.

Copied to clipboard

Yang Li, Tao Shen, Guodong Long, Jing Jiang, Tianyi Zhou, Chengqi Zhang

Information Extraction & Text Mining

Challenge:	Existing approaches to handle wrong labeling and long-tail relations are labor-intensive and scarce training data.
Approach:	They propose a neural network to handle wrong labeling and long-tail relations by collaborating relation-augmented attention.
Outcome:	The proposed neural network improves the state-of-the-art on the NYT dataset .

Superfiltering: Weak-to-Strong Data Filtering for Fast Instruction-Tuning (2024.acl-long)

Copied to clipboard

Ming Li, Yong Zhang, Shwai He, Zhitao Li, Hongyu Zhao, Jianzong Wang, Ning Cheng, Tianyi Zhou

Optimizing Length Compression in Large Reasoning Models (2026.acl-long)

Challenge:	Earlier studies of instruction tuning on Large Language Models focus on creating large, varied, and high-quality datasets with responses curated by human experts.
Approach:	They propose to use a smaller and weaker model to fine tune a larger and stronger model . they find it can largely speed up the data filtering and improve performance .
Outcome:	The proposed model can filter instruction data faster and better on benchmarks.

Copied to clipboard

Zhengxiang Cheng, Dongping Chen, Mingyang Fu, Tianyi Zhou

Contrastive Instruction Tuning (2024.findings-acl)

Challenge:	Large Reasoning Models suffer from producing unnecessary and verbose reasoning chains.
Approach:	They propose a post-training method that uses a Length Reward and a Compress Reward to remove the invalid portion of the thinking process.
Outcome:	The proposed method reduces sequence length by 50% with only a marginal (2%) drop in accuracy.

Copied to clipboard

Tianyi Yan, Fei Wang, James Y. Huang, Wenxuan Zhou, Fan Yin, Aram Galstyan, Wenpeng Yin, Muhao Chen

Tensorized Self-Attention: Efficiently Modeling Pairwise and Global Dependencies Together (N19-1)

Challenge:	Current LLMs exhibit limited robustness to unseen instructions, generating inconsistent outputs when the same instruction is phrased with slightly varied forms or language styles.
Approach:	They propose a method which maximizes the similarity between the hidden representations of semantically equivalent instruction-instance pairs while minimizing the similarities between semantically different ones.
Outcome:	Experiments on the PromptBench benchmark show that Contrastive Instruction Tuning improves LLMs’ robustness to unseen instructions with variations across character, word, sentence, and semantic levels by +2.5% in accuracy.

Copied to clipboard

Tao Shen, Tianyi Zhou, Guodong Long, Jing Jiang, Chengqi Zhang

Multi-Objective Linguistic Control of Large Language Models (2024.findings-acl)

Challenge:	Neural networks equipped with self-attention have parallelizable computation and the ability to capture both long-range and local dependencies.
Approach:	They propose a novel attention mechanism called "Multi-mask Tensorized Self-Attention" it captures pairwise and global dependencies by a compatibility function composed of dot-product and additive attentions .
Outcome:	The proposed model outperforms CNN-/RNN-/attention-based models on nine NLP benchmarks with compelling memory- and time-efficiency.

Copied to clipboard

Dang Nguyen, Jiuhai Chen, Tianyi Zhou

Retrieval-Augmented Retrieval: Large Language Models are Strong Zero-Shot Retriever (2024.findings-acl)

Challenge:	Existing Large language models prefer to generate verbose responses due to the length bias, which may increase unnecessary reading complexity.
Approach:	They propose to use off-the-shelf data to fine tune multiple linguistic complexities of LLM outputs to improve multi-complexity controllability and improve the quality of the responses.
Outcome:	The proposed method improves multi-complexity controllability significantly and retains or enhances the quality of the responses as a side benefit.

Copied to clipboard

Tao Shen, Guodong Long, Xiubo Geng, Chongyang Tao, Yibin Lei, Tianyi Zhou, Michael Blumenstein, Daxin Jiang

Eliminating Sentiment Bias for Aspect-Level Sentiment Classification with Unsupervised Opinion Extraction (2021.findings-emnlp)

Challenge:	Large-scale retrieval is indispensable in information-seeking tasks such as open-domain question answering and knowledgegrounded dialogue.
Approach:	They propose to use a large language model (LLM) to augment a query with its potential answers by prompting LLMs with a composition of the query and the query’s in-domain candidates.
Outcome:	The proposed method breaks brute-force combinations of retrievers with LLMs and lifts the performance of zero-shot retrieval to be very competitive on benchmark datasets.

Copied to clipboard

Bo Wang, Tao Shen, Guodong Long, Tianyi Zhou, Yi Chang

Sentiment Analysis

Challenge:	Aspect-level sentiment classification (ALSC) is a practical setting in aspect-based sentiment analysis due to no opinion term labeling needed, but it fails to interpret why a sentiment polarity is derived for the aspect.
Approach:	They propose a span-based anti-bias aspect representation learning framework that eliminates the sentiment bias in the aspect embedding by adversarial learning against aspects’ prior sentiment.
Outcome:	The proposed framework achieves state-of-the-art performance on five benchmarks, with the capability of unsupervised opinion extraction.

Where to show Demos in Your Prompt: A Positional Bias of In-Context Learning (2025.emnlp-main)

Copied to clipboard

Kwesi Adu Cobbina, Tianyi Zhou

Robust Natural Language Understanding with Residual Attention Debiasing (2023.findings-acl)

Challenge:	In-context learning (ICL) is a critical emerging capability of large language models (LLMs), enabling few-shot learning during inference by including a few demonstrations in the prompt.
Approach:	They propose to use positional bias to study ICL's performance for the first time by examining the positional variation in demos, system prompt, and user message in LLM input.
Outcome:	The proposed model can predict accuracy and accuracy when demos are placed at different positions in the input prompt and in the user message.

Copied to clipboard

Fei Wang, James Y. Huang, Tianyi Yan, Wenxuan Zhou, Muhao Chen

Robustness in NLP

Challenge:	Existing ensemble-based debiasing methods do not address unintended dataset biases . attention plays a crucial role in providing robust prediction in NLU models .
Approach:	They propose an end-to-end debiasing method that mitigates unintended biases from attention.
Outcome:	The proposed method improves the OOD performance of BERT-based models on three benchmarks.

Understanding the Thinking Process of Reasoning Models: A Perspective from Schoenfeld’s Episode Theory (2025.emnlp-main)

Copied to clipboard

Ming Li, Nan Zhang, Chenrui Fan, Hong Jiao, Yanbin Fu, Sydney Peters, Qingshu Xu, Robert Lissitz, Tianyi Zhou

Linguistics & Cognitive NLP

Challenge:	Large Reasoning Models (LRMs) generate extensive chain-of-thought reasoning, but we lack a principled framework for understanding how these thoughts are structured.
Approach:	They propose a method to analyze the reasoning traces of Large Reasoning Models using Schoenfeld’s Episode Theory.
Outcome:	The proposed framework provides a theoretically grounded methodology for interpreting LRM cognition and enables future work on more controllable and transparent reasoning systems.

AI, Take the Wheel: What Drives Delegation and Trust in Human–Computer Cooperative Question Answering? (2026.findings-acl)

Copied to clipboard

Maharshi Gor, Yoo Yeon Sung, Yu Hou, Eve Fleisig, Zhu Irene Ying, Tianyi Zhou, Jordan Lee Boyd-Graber

Natural Language Interfaces

Challenge:	Human-AI collaboration is already happening, both in proactive delegation and deliberative adoption settings.
Approach:	They study delegating a task to AI without seeing its output and evaluating AI suggestions to decide whether to adopt them how AI output shapes final decisions.
Outcome:	The proposed game pairs 23 experts with 16 AI agents, capturing 387 delegation and 1440 adoption decisions.

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.

Can LLMs Estimate Student Struggles? Human-AI Difficulty Alignment with Proficiency Simulation for Item Difficulty Prediction (2026.findings-acl)

Copied to clipboard

Ming Li, Han Chen, Yunze Xiao, Jian Chen, Hong Jiao, Tianyi Zhou

Cognitive Modeling

Challenge:	Accurate estimation of item (question or task) difficulty suffers from the cold start problem.
Approach:	They propose to use large-scale empirical analysis to examine human-AI Difficulty Alignment . they find that models struggle to simulate the capability limitations of students .
Outcome:	The proposed model size is not reliably helpful for human-AI alignment . high performance often impedes accurate difficulty estimation, the authors say .

LLMBox: A Comprehensive Library for Large Language Models (2024.acl-demos)

Copied to clipboard

Tianyi Tang, Hu Yiwen, Bingqian Li, Wenyang Luo, ZiJing Qin, Haoxiang Sun, Jiapeng Wang, Shiyi Xu, Xiaoxue Cheng, Geyang Guo, Han Peng, Bowen Zheng, Yiru Tang, Yingqian Min, Yushuo Chen, Jie Chen, Ranchi Zhao, Luran Ding, Yuhao Wang, Zican Dong, Xia Chunxuan, Junyi Li, Kun Zhou, Xin Zhao, Ji-Rong Wen

A Usage-centric Take on Intent Understanding in E-Commerce (2024.emnlp-main)

Challenge:	a library to facilitate the development, use, and evaluation of large language models (LLMs) is presented.
Approach:	They propose a unified library to facilitate the development, use and evaluation of large language models (LLMs).
Outcome:	The proposed library is based on extensive experiments in a variety of evaluation settings.

Copied to clipboard

Wendi Zhou, Tianyi Li, Pavlos Vougiouklis, Mark Steedman, Jeff Pan

Challenge:	Identifying and understanding user intents is a crucial task for E-Commerce.
Approach:	They propose to use intent understanding as a natural language reasoning task independent of product ontologies to identify and understand user intents.
Outcome:	The proposed framework can't be used to strongly align user intents with products with desirable properties and recommend useful products across diverse categories.

Trait Activation in Silicon: A Situation-Aware Framework for Psychologically Grounded Role-Playing (2026.acl-long)

Copied to clipboard

Zuolong Li, Pingyu Wu, Xianwen Huang, Tianyi Wei, Wenbo Zhou

What Happened in LLMs Layers when Trained for Fast vs. Slow Thinking: A Gradient Perspective (2025.acl-long)

Challenge:	Role-playing agents lack a deep understanding of complex human psychological mechanisms.
Approach:	They propose a situation-aware framework that decouples personality traits into bidirectional LoRA adapters.
Outcome:	Empirical results show that PD-LLM achieves superior performance in both static fidelity and dynamic adaptability.

Copied to clipboard

Ming Li, Yanhong Li, Tianyi Zhou

Phrase-level Textual Adversarial Attack with Label Preservation (2022.findings-naacl)

Challenge:	Xu et al., 2024) study shows that slow thinking can distinguish correct and irrelevant reasoning paths.
Approach:	They investigate how fast vs. slow thinking affects layer-wise gradients in large language models . they find that slow thinking can distinguish correct and irrelevant reasoning paths .
Outcome:	The results show that slow thinking can distinguish correct and irrelevant reasoning paths.

Copied to clipboard

Yibin Lei, Yu Cao, Dianqi Li, Tianyi Zhou, Meng Fang, Mykola Pechenizkiy

Robustness in NLP

Challenge:	Existing adversarial attacks are usually realized through word-level or sentence-level perturbations, which either limit the perturbation space or sacrifice fluency and textual quality.
Approach:	They propose a phrase-level perturbation-based adversarial ATtack that generates adversarials through phrase- level perturbations.
Outcome:	The proposed approach improves the performance of natural language processing models by reducing the need for word-level perturbations and preserving the fluency and grammaticality of the samples.

Wait, We Don’t Need to “Wait”! Removing Thinking Tokens Improves Reasoning Efficiency (2025.findings-emnlp)

Copied to clipboard

Chenlong Wang, Yuanning Feng, Dongping Chen, Zhaoyang Chu, Ranjay Krishna, Tianyi Zhou

Enhancing Visual-Language Modality Alignment in Large Vision Language Models via Self-Improvement (2025.findings-naacl)

Challenge:	Recent advances in large reasoning models often introduce significant overthinking . this leads to verbose and redundant outputs that hinder efficiency.
Approach:	They propose a plug-and-play solution that disables explicit self-reflection . it suppresses tokens such as "Wait" and "Hmm" during inference .
Outcome:	The proposed approach reduces chain-of-thought trajectory length by up to 27%–51% in five R1-style model series without compromising model utility.

Copied to clipboard

Xiyao Wang, Jiuhai Chen, Zhaoyang Wang, Yuhang Zhou, Yiyang Zhou, Huaxiu Yao, Tianyi Zhou, Tom Goldstein, Parminder Bhatia, Taha Kass-Hout, Furong Huang, Cao Xiao