Papers by Shu Zhao
The Impact of Reasoning Step Length on Large Language Models (2024.findings-acl)
Copied to clipboard
| Challenge: | Long reasoning steps in LLMs improve reasoning abilities, but the correlation between their effectiveness and the length of reasoning steps remains largely unknown. |
| Approach: | They conducted experiments that expand and compress the rationale reasoning steps within CoT demonstrations while keeping all other factors constant. |
| Outcome: | The results show that lengthening the reasoning steps in prompts significantly enhances LLMs’ reasoning abilities across multiple datasets. |
Enhancing Extractive Question Answering in Multiparty Dialogues with Logical Inference Memory Network (2025.coling-main)
Copied to clipboard
| Challenge: | Existing models for multiparty dialogue question answering (QA) do not consider logical inference relations in multiparty dialogs, leading to suboptimal performance. |
| Approach: | They propose a memory network with logical inference for extractive QA in multiparty dialogues. |
| Outcome: | The proposed model achieves state-of-the-art on Molweni and FriendsQA benchmarks. |
Are Intermediate Layers and Labels Really Necessary? A General Language Model Distillation Method (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing knowledge distillation methods rely on intermediate layer features and golden labels, which require aligned model architecture and labeled data respectively. |
| Approach: | They propose a general language model distillation method that performs two-stage word prediction distillation and vocabulary compression, which is simple and shows extremely strong performance. |
| Outcome: | The proposed method outperforms 25 state-of-the-art methods on the SuperGLUE benchmark, achieving an average score that surpasses the best method by 3%. |
Rethinking Composed Image Retrieval Evaluation: A Fine-Grained Benchmark from Image Editing (2026.acl-long)
Copied to clipboard
Tingyu Song, Yanzhao Zhang, Mingxin Li, Zhuoning Guo, Dingkun Long, Pengjun Xie, Siyue Zhang, Yilun Zhao, Shu Wu
| Challenge: | Composed Image Retrieval (CIR) is a complex task in multimodal understanding . current CIR benchmarks lack a robust evaluation pipeline and limited query categories . |
| Approach: | They construct a fine-grained CIR benchmark that allows for precise control over modification types and content. |
| Outcome: | The proposed benchmark covers 5,000 high-quality queries structured across five main categories and fifteen subcategories. |
Explore the Reasoning Capability of LLMs in the Chess Testbed (2025.naacl-short)
Copied to clipboard
| Challenge: | a recent study shows that large language models struggle with long-term, complex reasoning tasks. |
| Approach: | They propose to integrate annotated strategy and tactic into large language models to improve reasoning capability. |
| Outcome: | The proposed model performs better than GPT, Claude, and Gemini models . it integrates annotated strategy and tactic into the model . |
MTVQA: Benchmarking Multilingual Text-Centric Visual Question Answering (2025.findings-acl)
Copied to clipboard
Jingqun Tang, Qi Liu, Yongjie Ye, Jinghui Lu, Shu Wei, An-Lan Wang, Chunhui Lin, Hao Feng, Zhen Zhao, Yanjie Wang, Yuliang Liu, Hao Liu, Xiang Bai, Can Huang
| Challenge: | Text-Centric Visual Question Answering (TEC-VQA) is a text-centric visual task understanding tool. |
| Approach: | They introduce a benchmark that features human expert annotations across 9 languages . they prioritize the text in question-answer pairs while disregarding visual text in images . |
| Outcome: | The proposed benchmarks prioritize the text in question-answer pairs while disregarding visual text in images. |
Forget the Unneeded: Backdooring Large Language Models via Contrastive-enhanced Machine Unlearning (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for prompt tuning for Large Language Models find backdoor attacks to be significant in data-rich scenarios. |
| Approach: | They propose a backdoor attacks through contrastive-enhanced machine unlearning in data-limited scenarios . they use a machine un learning method to capture precise backdoor patterns . |
| Outcome: | The proposed method captures precise backdoor patterns without association between triggers and backdoors, reducing side effects. |
A Survey on Sparse Autoencoders: Interpreting the Internal Mechanisms of Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Sparse Autoencoders (SAEs) can disentangle complex features into more interpretable components. |
| Approach: | They propose to use Sparse Autoencoders to disentangle LLM features into more interpretable components. |
| Outcome: | The proposed method disentangles complex features into more interpretable components. |
RPC-Bench: A Fine-grained Benchmark for Research Paper Comprehension (2026.acl-long)
Copied to clipboard
Yelin Chen, Fanjin Zhang, Suping Sun, Yunhe Pang, Yuanchun Wang, Jian Song, XiaoYan Li, Lei Hou, Shu Zhao, Jie Tang, Juanzi Li
| Challenge: | Existing benchmarks for understanding research papers offer limited fine-grained evaluation at scale. |
| Approach: | They propose a large-scale question-answering benchmark built from review–rebuttal exchanges of high-quality computer science papers. |
| Outcome: | The proposed model is based on human-verified QA pairs and contains 15K questions. |
Generate First, Then Sample: Enhancing Fake News Detection with LLM-Augmented Reinforced Sampling (2025.acl-long)
Copied to clipboard
| Challenge: | Existing models have a performance gap of 20% between classifying fake news and real news, making them less suitable for practical deployment. |
| Approach: | They propose to adopt an LLM to generate fake news in three different styles, which are later incorporated into the training set to augment the representation of fake news. |
| Outcome: | The proposed model achieves state-of-the-art performance on two benchmark datasets and improves detection accuracy by 24.02% and 11.06% respectively. |
Enhancing Text-to-SQL Capabilities of Large Language Models through Tailored Promptings (2024.lrec-main)
Copied to clipboard
| Challenge: | Large language models with prompting have achieved encouraging results on many natural language processing tasks due to the absence of task-tailored promptings. |
| Approach: | They propose three promptings specifically designed for Text-to-SQL: SL-prompt, CC-promped, and SL+CC prompt. |
| Outcome: | The proposed promptings achieve execution accuracy of 86.2% and test-suite accuracy of 76% . the granularity of schema linking and the order of clause generation have great impact on performance, which are considered little in previous research. |
Large Vision-Language Model Alignment and Misalignment: A Survey Through the Lens of Explainability (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Large Vision-Language Models have demonstrated remarkable capabilities in processing both visual and textual information. |
| Approach: | They examine the challenge of alignment and misalignment in LVLMs through an explainability lens. |
| Outcome: | The findings highlight the need for standardized evaluation protocols and in-depth explainability studies. |
From Generation to Judgment: Opportunities and Challenges of LLM-as-a-judge (2025.emnlp-main)
Copied to clipboard
Dawei Li, Bohan Jiang, Liangjie Huang, Alimohammad Beigi, Chengshuai Zhao, Zhen Tan, Amrita Bhattacharjee, Yuxuan Jiang, Canyu Chen, Tianhao Wu, Kai Shu, Lu Cheng, Huan Liu
| Challenge: | Recent advances in Large Language Models (LLMs) inspire the "LLM-as-a-judge" paradigm . traditional methods of assessment and evaluation fail in dynamic and open-ended scenarios . |
| Approach: | They propose a paradigm where LLMs are leveraged to perform scoring, ranking, or selection for machine learning evaluation scenarios. |
| Outcome: | The proposed model-based judgment and evaluation paradigms are based on large language models and are compared to the current model-driven evaluation paradigm. |
Let’s Be Self-generated via Step by Step: A Curriculum Learning Approach to Automated Reasoning with Large Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing efforts to improve CoT prompting have limitations that require extensive human effort or performance needs to be improved. |
| Approach: | They propose a prompt approach for automatic reasoning called LBS3 inspired by curriculum learning which better reflects human learning habits. |
| Outcome: | The proposed approach achieves strongly competitive performance compared to baselines in reasoning-intensive tasks with varying open- and closed-source LLMs. |
PDALN: Progressive Domain Adaptation over a Pre-trained Model for Low-Resource Cross-Domain Named Entity Recognition (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches to Named Entity Recognition (NER) are limited in labeled resources and domain shift. |
| Approach: | They propose a progressive domain adaptation knowledge distillation approach to adapt high-resource domains to low-resourced target domains by employing three components to achieve superior domain adaptability. |
| Outcome: | The proposed approach can adapt high-resource domains to low-resourced target domains even if they are diverse in terms and writing styles. |
Seeking Common but Distinguishing Difference, A Joint Aspect-based Sentiment Analysis Model (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing models focus on aspect term extraction, opinion term extraction and sentiment polarity classification but ignore the difference. |
| Approach: | They propose a joint aspect-based sentiment analysis task that focuses on the difference between the two tasks to improve the model's robustness. |
| Outcome: | Empirical results show that the proposed model outperforms the previous state-of-the-art on four benchmark datasets. |
Rejection-to-Acceptance Transition: Model Editing-Based Jailbreak Backdoor Injection Not Limited to Few Output Tokens (2026.findings-acl)
Copied to clipboard
Shiji Yang, Min Cai, Hao Xiong, Congyao Mei, Haodong Zou, Shicheng Tan, Jie Chen, Fulan Qian, Shu Zhao
| Challenge: | Existing methods for jailbreaking LLMs are implemented by binding backdoors to predefined phrases as first few output tokens, inducing the LLM’s next-token prediction to produce continuous responses. |
| Approach: | They propose a model editing-based jailbreak backdoor attack that hijacks LLM representations into a acceptance domain rather than binding to a few output tokens. |
| Outcome: | The proposed model editing method outperforms existing methods, showing stronger jailbreak capabilities across LLMs and datasets. |
TableLLM: Enabling Tabular Data Manipulation by LLMs in Real Office Usage Scenarios (2025.findings-acl)
Copied to clipboard
Xiaokang Zhang, Sijia Luo, Bohan Zhang, Zeyao Ma, Jing Zhang, Yang Li, Guanlin Li, Zijun Yao, Kangli Xu, Jinchang Zhou, Daniel Zhang-Li, Jifan Yu, Shu Zhao, Juanzi Li, Jie Tang
| Challenge: | TableLLM is a robust large language model capable of handling tabular data manipulation tasks. |
| Approach: | They propose a distant supervision method for training which includes a reasoning process extension strategy and a cross-way validation strategy. |
| Outcome: | The proposed model has 8 billion parameters and is capable of handling tabular data tasks. |
FinDABench: Benchmarking Financial Data Analysis Ability of Large Language Models (2025.coling-main)
Copied to clipboard
Shu Liu, Shangqing Zhao, Chenghao Jia, Xinlin Zhuang, Zhaoguang Long, Jie Zhou, Aimin Zhou, Man Lan, Yang Chong
| Challenge: | Large Language Models (LLMs) have demonstrated impressive capabilities across a wide range of tasks, but their proficiency and reliability in the specialized domain of financial data analysis remain uncertain. |
| Approach: | FinDABench is a benchmark designed to evaluate the financial data analysis capabilities of Large Language Models (LLMs) it comprises 15,200 training instances and 8,900 test instances, all meticulously crafted by human experts. |
| Outcome: | FinDABench measures the financial data analysis capabilities of large language models (LLMs) across three dimensions: 1) Core Ability; 2) Analytical Ability; 3) Technical Ability. |
Beyond Input Activations: Identifying Influential Latents by Gradient Sparse Autoencoders (2025.emnlp-main)
Copied to clipboard
| Challenge: | Sparse Autoencoders (SAEs) have recently emerged as powerful tools for interpreting and steering the internal representations of large language models (LLMs). |
| Approach: | They propose a method that identifies the most influential latents by incorporating output-side gradient information. |
| Outcome: | The proposed method identifies the most influential latents by incorporating output-side gradient information. |
Execution as Verification: Fine-Grained Self-Correcting Reasoning for Complex KBQA (2026.acl-long)
Copied to clipboard
| Challenge: | Existing knowledge base question answering methods are limited by syntactic constraints and are prone to structural deviations that render queries unexecutable. |
| Approach: | They propose a framework that reframes semantic parsing as an iterative reasoning process driven by execution feedback. |
| Outcome: | The proposed method achieves significant improvements in query executability and answer accuracy on the WebQSP and CWQ datasets. |
REACT: Representation Extraction And Controllable Tuning to Overcome Overfitting in LLM Knowledge Editing (2025.emnlp-main)
Copied to clipboard
Haitian Zhong, Yuhuan Liu, Ziyang Xu, Guofan Liu, Qiang Liu, Shu Wu, Zhe Zhao, Liang Wang, Tieniu Tan
| Challenge: | Large language model editing methods suffer from overfitting, where factual updates can propagate beyond their intended scope, overemphasizing the edited target even when it’s contextually inappropriate. |
| Approach: | They propose a framework for precise and controllable knowledge editing that utilizes two-phase representations and a linear transformation to compute a directional "belief shift" vector. |
| Outcome: | The proposed framework significantly reduces overfitting across nearly all evaluation metrics and on COUNTERFACT and MQuAKE. |
GKD: A General Knowledge Distillation Framework for Large-scale Pre-trained Language Model (2023.acl-industry)
Copied to clipboard
| Challenge: | Existing knowledge distillation frameworks for language models are limited by memory and the use of complex distillation methods on larger-scale PLMs. |
| Approach: | They propose a general knowledge distillation framework that supports distillation on larger-scale PLMs using various distillation methods. |
| Outcome: | The proposed framework can support distillation on larger-scale PLMs and 25 mainstream methods on 8 NVIDIA A100 (40GB) GPUs. |
From Imitation to Introspection: Probing Self-Consciousness in Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing language models demonstrate impressive abilities in areas like natural language understanding, content creation, and reasoning. |
| Approach: | They propose a definition of self-consciousness for language models and refine ten core concepts by leveraging structural causal games. |
| Outcome: | The proposed definitions are based on structural causal games and ten core concepts. |
CompTab: A Comprehensive Benchmark for Real-World TableQA with Complex Reasoning and Irregular Tables (2026.acl-long)
Copied to clipboard
Zhen Yang, Wei Du, Jie Wang, Wenze Zhou, Xiangfeng Meng, Zhengyang Wang, Suping Sun, Ziwei Du, Haodong Zou, Jie Chen, Yongbin Liu, Shicheng Tan, Jiahao Ying, Shu Zhao
| Challenge: | Existing benchmarks focus on well-structured tables and fail to reflect irregular structures and complex reasoning commonly encountered in real-world scenarios. |
| Approach: | They propose a benchmark to evaluate TableQA under complex reasoning and irregular table conditions. |
| Outcome: | The proposed framework improves generalization and realism of large language models under complex and irregular table conditions. |
Thinking Beyond the Local: Multi-View Instructed Adaptive Reasoning in KG-Enhanced LLMs (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing methods for large language models adopt query-driven iterative reasoning from a local perspective, limiting efficiency and accuracy for complex multi-hop tasks. |
| Approach: | They propose a multi-view instructed adaptive reasoning of LLM on Knowledge Graphs that allows LLMs to plan, evaluate, and adapt reasoning paths from a global perspective. |
| Outcome: | The proposed model overcomes the limitations of local exploration by enabling LLMs to plan, evaluate, and adapt reasoning paths from a global perspective. |
RLShield: Dynamic Jailbreak Detection for LLMs via Reinforced Adaptive Learning (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to detect jailbreak prompts rely on static model components or fixed decision thresholds. |
| Approach: | They propose a dynamic jailbreak detection framework that employs reinforcement learning for adaptive threshold selection. |
| Outcome: | Experimental results show that the framework outperforms baselines in detection performance while maintaining high computational efficiency. |