Papers by Sirui Han

Omni-RewardBench: Toward a Comprehensive Evaluation of Generative Reward Models Across Modalities (2026.acl-long)

Challenge:	Existing research to improve CoT efficiency falls into three categories, each with distinct limitations.
Approach:	They propose a training-free framework that addresses both dimensions of CoT reasoning by applying a progressive precision reduction strategy coupled with an entropy-based confidence mechanism for adaptive termination.
Outcome:	Empirical results show that the proposed framework achieves 11.3 efficiency gain without compromising accuracy.

Copied to clipboard

Chi-Min Chan, Yujin Zhou, Pengcheng Wen, Boqin Yin, Jiaming Ji, Juntao Dai, Wei Xue, Sirui Han, Yike Guo

Challenge:	Existing evaluation benchmarks for ORMs are largely text-centric or limited to bimodal tasks . a new study examines the effectiveness of Omni-RewardBench for ORms across modalities .
Approach:	They propose a hybrid automatic-annotation and human-verification pipeline to construct high-quality evaluation data.
Outcome:	The proposed model is the first benchmark for comprehensive evaluation of ORMs across modalities.

ARise: Towards Knowledge-Augmented Reasoning via Risk-Adaptive Search (2025.acl-long)

Copied to clipboard

Yize Zhang, Tianshu Wang, Sirui Chen, Kun Wang, Xingyu Zeng, Hongyu Lin, Xianpei Han, Le Sun, Chaochao Lu

Reasoning

Challenge:	Large language models (LLMs) have impressive capabilities but their application in open-ended, knowledge-intensive, complex reasoning scenarios is limited.
Approach:	They propose a framework that integrates risk assessment of intermediate reasoning states with dynamic retrieval-augmented generation within a Monte Carlo tree search paradigm.
Outcome:	The proposed framework outperforms the state-of-the-art KAR methods by up to 23.10% and the latest RAG-equipped large reasoning models by upto 25.37%.

Towards Advanced Mathematical Reasoning for LLMs via First-Order Logic Theorem Proving (2025.emnlp-main)

Copied to clipboard

Chuxue Cao, Mengze Li, Juntao Dai, Jinluan Yang, Zijian Zhao, Shengyu Zhang, Weijie Shi, Chengzhong Liu, Sirui Han, Yike Guo

Reasoning

Challenge:	Large language models (LLMs) have shown promising first-order logic (FOL) reasoning capabilities with applications in various areas, but their effectiveness in complex mathematical reasoning involving multi-step FOL deductions remains under-explored.
Approach:	They propose a self-adaptive solution that enhances the Diversity and REAsonability of LLMs’ generation strategies by introducing an Axiom-Driven Strategy Diversification mechanism and a Sub-Proposition Error Feedback to help LLM reflect on and correct their proofs.
Outcome:	The proposed model improves diversity and REAsonability of LLMs’ generation strategies by introducing an Axiom-Driven Strategy Diversification mechanism and a Sub-Proposition Error Feedback to help LLM reflect on and correct proofs.

Out-of-Distribution Detection via LLM-Guided Outlier Generation for Text-attributed Graph (2025.findings-acl)

Copied to clipboard

Xiangwei Lv, Mengze Li, Jingyuan Chen, Zhiang Dong, Sirui Han, Beishui Liao

Context Reasoner: Incentivizing Reasoning Capability for Contextualized Privacy and Safety Compliance via Reinforcement Learning (2025.emnlp-main)

Challenge:	Text-Attributed Graphs (TAGs) are widely used in the real world.
Approach:	They propose to use Large Language Models to generate OOD-nodes with high quality . they also use LLMs to integrate existing nodes with LLM-generated edges .
Outcome:	The proposed method performs well on samples outside the In-Distribution (ID) data, but it is difficult to obtain high-quality OOD samples in the real world.

Copied to clipboard

Wenbin Hu, Haoran Li, Huihao Jing, Qi Hu, Ziqian Zeng, Sirui Han, Xu Heli, Tianshu Chu, Peizhao Hu, Yangqiu Song

BTC-LLM: Efficient Sub-1-Bit LLM Quantization via Learnable Transformation and Binary Codebook (2026.acl-long)

Challenge:	Current mitigation strategies fail to preserve contextual reasoning capabilities in risky scenarios, leading to systemic risks for legal compliance.
Approach:	They propose to use reinforcement learning with a rule-based reward to incentivize contextual reasoning capabilities while enhancing compliance with safety and privacy norms.
Outcome:	The proposed model outperforms Qwen2.5-7B-Instruct model in safety and privacy benchmarks and achieves +8.58% accuracy improvement.

Copied to clipboard

Hao Gu, Lujun Li, Hao Wang, Lei Wang, Zheyu Wang, Bei Liu, Jiacheng Liu, Qiyuan Zhu, Sirui Han, Yike Guo

Text Clustering

Challenge:	Recent sparsity-aware binarization approaches can achieve sub-1-bit compression, but they face performance degradation, mask-management overhead, and limited hardware compatibility.
Approach:	They propose a binary quantization framework that leverages binary pattern clustering and weight transformation to overcome performance degradation and mask-management overhead.
Outcome:	The proposed framework achieves state-of-the-art compression (1.11–0.7 bits) it maintains high performance with only a 3.1% accuracy drop in zero-shot benchmarks while delivering a 1.6 speedup over FP16.

Perception, Understanding and Reasoning: A Multimodal Benchmark for Video Fake News Detection (2026.acl-long)

Copied to clipboard

Cui Yakun, Peng Qi, Fushuo Huo, Hang Du, Weijie Shi, Juntao Dai, Zhenghao Zhu, Sirui Han

InsightEval: An Expert-Curated Benchmark for Assessing Insight Discovery in LLM-Driven Data Agents (2026.findings-acl)

Challenge:	Existing video fake news detection benchmarks focus on the detection accuracy, while failing to provide fine-grained assessments for the entire detection process.
Approach:	They propose a process-oriented video fake news detection benchmark that evaluates MLLMs' perception, understanding, and reasoning capabilities in VFND.
Outcome:	The proposed model achieves sota performance on video fake news detection tasks.

Copied to clipboard

Zhenghao Zhu, Yuanfeng Song, Xing Chen, Chengzhong Liu, Cui Yakun, Caleb Chen Cao, Sirui Han, Yike Guo

Challenge:	Existing frameworks for data analysis and insight exploration are lacking in terms of benchmarks . existing frameworks suffer from format inconsistencies, poorly conceived objectives, and redundant insights.
Approach:	They propose a data-curation pipeline to construct a new dataset named InsightEval.
Outcome:	The proposed benchmarks highlight prevailing challenges in automated insight discovery and raise key findings to guide future research.

DIDS: Domain Impact-aware Data Sampling for Large Language Model Training (2025.emnlp-main)

Copied to clipboard

Weijie Shi, Jipeng Zhang, Yaguang Wu, Jingzhi Fang, Shibo Zhang, Yao Zhao, Hao Chen, Ruiyuan Zhang, Yue Cui, Jia Zhu, Sirui Han, Jiajie Xu, Xiaofang Zhou

Not All Contexts Are Equal: Teaching LLMs Credibility-aware Generation (2024.emnlp-main)

Challenge:	Existing approaches for optimizing domain-level sampling strategies struggle with maintaining intra-domain consistency and accurately measuring domain impact.
Approach:	They propose to use a Fisher-Information Matrix-guided metric to measure domain impact to ensure intra-domain consistency and accuracy.
Outcome:	The proposed model achieves 3.4% higher average performance while maintaining comparable training efficiency.

Copied to clipboard

Ruotong Pan, Boxi Cao, Hongyu Lin, Xianpei Han, Jia Zheng, Sirui Wang, Xunliang Cai, Le Sun

Information Retrieval

Challenge:	Existing RAG paradigms suffer from the impact of flawed information introduced during the retrieval phrase, thereby diminishing the reliability and correctness of the generated output.
Approach:	They propose a framework that empowers models to discern and process information based on its credibility.
Outcome:	The proposed framework outperforms existing models with retrieval augmentation and exhibits robustness despite increasing noise in the context.

Learning While Staying Curious: Entropy-Preserving Supervised Fine-Tuning via Adaptive Self-Distillation for Large Reasoning Models (2026.acl-long)

Copied to clipboard

Hao Wang, Hao Gu, Hongming Piao, Kaixiong Gong, Yuxiao Ye, Xiangyu Yue, Sirui Han, Yike Guo, Dapeng Wu

Bit-by-Bit: Progressive QAT Strategy with Outlier Channel Splitting for Stable Low-Bit LLMs (2026.acl-long)

Challenge:	Recent advances establish "SFT-then-RL" as the defacto paradigm for enhancing large reasoning mod- els on automatically verifiable tasks.
Approach:	They propose an entropy-preserving SFT method to enhance exploration capabilities through intrinsic curiosity.
Outcome:	The proposed method outperforms the vanilla method on reasoning tasks by 2.5 points . it also outperformed the vanilla SFT by 2.9 points on out-of-distribution tasks .

Copied to clipboard

Binxing Xu, Hao Gu, Lujun Li, Hao Wang, Bei Liu, Jiacheng Liu, Qiyuan Zhu, Xintong Yang, Chao Li, Sirui Han, Yike Guo

Green & Sustainable NLP

Challenge:	Existing approaches to training LLMs at ultra-low precisions suffer from convergence instability and substantial training costs.
Approach:	They propose a progressive QAT framework with outlier channel splitting to address these issues . they use nested structure of integer quantization grids to enable a "train once, deploy any precision" paradigm .
Outcome:	The proposed framework outperforms baselines on both Llama2/3 and W2A16, with an 11 speedup over BF16.

VizoMem: A Visual-Textual Memory Framework for Efficient Long-Horizon Reasoning (2026.findings-acl)

Copied to clipboard

Weijie Liang, Yuanfeng Song, Xing Chen, Caleb Chen Cao, Sirui Han, Yike Guo

Visual Data in NLP

Challenge:	Existing systems that use long-context modeling incur computational and memory overhead.
Approach:	They propose a visual memory framework that pre-rendered text into structured images and stored as visual notes for agentic systems.
Outcome:	The proposed system reduces token consumption while preserving effective long-term memory recall.

SafeMT: Multi-turn Safety for Multimodal Language Models (2026.acl-long)

Copied to clipboard

Han Zhu, Juntao Dai, Jiaming Ji, Haoran Li, Chengkun Cai, Pengcheng Wen, Chi-Min Chan, Boyuan Chen, Yaodong Yang, Sirui Han, Yike Guo

Glance-or-Gaze: Incentivizing LMMs to Adaptively Focus Search via Reinforcement Learning (2026.findings-acl)

Challenge:	Multi-turn dialogues pose a greater risk than single prompts, but existing safety benchmarks do not account for this situation.
Approach:	They propose a benchmark that features dialogues of varying lengths generated from harmful queries accompanied by images.
Outcome:	The proposed model reduces multi-turn Attack Success Rate (ASR) compared to existing guard models.

Copied to clipboard

Hongbo Bai, Yujin Zhou, Yile Wu, Chi-Min Chan, Pengcheng Wen, Kunhao Pan, Sirui Han, Yike Guo

Boosting Policy and Process Reward Models with Monte Carlo Tree Search in Open-Domain QA (2025.findings-acl)

Challenge:	Existing search-augmented approaches rely on indiscriminate whole-image retrieval and lack deep iterative reflection, limiting their effectiveness on complex visual queries.
Approach:	They propose a fully autonomous framework that shifts from passive perception to active visual planning and introduces a Selective Gaze mechanism that dynamically chooses whether to glance at global context or gaze into high-value regions.
Outcome:	Experiments across six benchmarks demonstrate state-of-the-art performance.

Copied to clipboard

Chi-Min Chan, Chunpu Xu, Junqi Zhu, Jiaming Ji, Donghai Hong, Pengcheng Wen, Chunyang Jiang, Zhen Ye, Yaodong Yang, Wei Xue, Sirui Han, Yike Guo

Question Answering

Challenge:	Experimental results show that our approach can effectively improve the performance of both the policy model and the reward model.
Approach:	They propose to use Monte Carlo Tree Search for both policy model improvement and reward model improvement to bridge it to more subtle open-domain question answering.
Outcome:	The proposed approach surpasses existing methods for annotation and training data with fewer data points and achieves better performance in test-time scaling strategies.

Benchmarking Fine-Grained Error Detection in Multimodal Reasoning (2026.acl-long)

Copied to clipboard

Chi-Min Chan, Han Zhu, Chunyang Jiang, Jiaming Ji, Juntao Dai, Wei Xue, Sirui Han, Yike Guo

RAG over Tables: Hierarchical Memory Index, Multi-Stage Retrieval, and Benchmarking (2026.findings-acl)

Challenge:	Multimodal Process Reward Models (MPRMs) have emerged as a pivotal framework for enhancing the reasoning capabilities of Multimodal Large Language Models.
Approach:	They propose a benchmark specifically designed to evaluate MPRMs’ proficiency in detecting erroneous reasoning steps across diverse error categories.
Outcome:	The proposed model achieves up to 4.8% performance improvement through test-time scaling.

Copied to clipboard

Jiaru Zou, Dongqi Fu, Sirui Chen, Xinrui He, Zihao Li, Yada Zhu, Jiawei Han, Jingrui He

Structured Data in NLP

Challenge:	Retrieval-Augmented Generation (RAG) integrates knowledge from tables with an external knowledge base to improve the answer relevance and accuracy.
Approach:	They propose a table-corpora-aware RAG framework called T-RAG to integrate external knowledge into Large Language Models (LLMs) they then develop a multi-table question answering benchmark called MultiTableQA which spans 3 different task types, 57,193 tables, and 23,758 questions in total.
Outcome:	The proposed framework achieves state-of-the-art accuracy, recall, and runtime performance, with improvements of up to 9.4%.

Automate Strategy Finding with LLM in Quant Investment (2025.findings-emnlp)

Copied to clipboard

Zhizhuo Kou, Holam Yu, Junyu Luo, Jingshu Peng, Xujia Li, Chengzhong Liu, Juntao Dai, Lei Chen, Sirui Han, Yike Guo

QaRL: Rollout-Aligned Quantization-Aware RL for Fast and Stable Training under Training–Inference Mismatch (2026.findings-acl)

Challenge:	Experimental results demonstrate robust performance of the strategy in Chinese & US market regimes compared to established benchmarks.
Approach:	They propose a framework leveraging Large Language Models within a risk-aware multi-agent system for automate strategy finding in quantitative finance.
Outcome:	The proposed framework outperforms all benchmarks in Chinese & US market regimes with 53.17% cumulative return on SSE50.

Copied to clipboard

Hao Gu, Hao Wang, Jiacheng Liu, Lujun Li, Qiyuan Zhu, Bei Liu, Binxing Xu, Lei Wang, Xintong Yang, Sida Lin, Sirui Han, Yike Guo

Challenge:	Recent work has shown that reinforcement learning with simple rule-based reward functions (RLVR) can induce emergent reasoning behaviors and yield gains in challenging domains such as math problem solving.
Approach:	They propose a rollout-alignment-quantization-aware RL which aligns training-side forward with the quantized rollout to minimize mismatch.
Outcome:	The proposed approach outperforms quantized-rollout training by +5.5 on Qwen3-30B-A3B MoE for math problems while maintaining low-bit throughput.

LegalReasoner: Step-wised Verification-Correction for Legal Judgment Reasoning (2025.acl-long)

Copied to clipboard

Weijie Shi, Han Zhu, Jiaming Ji, Mengze Li, Jipeng Zhang, Ruiyuan Zhang, Jia Zhu, Jiajie Xu, Sirui Han, Yike Guo

Reasoning

Challenge:	Existing legal judgment prediction methods struggle with logical errors when conducting complex legal reasoning.
Approach:	They propose a method which enhances LJP reliability through step-wise verification and correction of the reasoning process.
Outcome:	The proposed model significantly improves concordance with court decisions from 72.37 to 80.27 on LLAMA-3.1-70B.

PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preference (2025.acl-long)

Copied to clipboard

Jiaming Ji, Donghai Hong, Borong Zhang, Boyuan Chen, Josef Dai, Boren Zheng, Tianyi Alex Qiu, Jiayi Zhou, Kaile Wang, Boxun Li, Sirui Han, Yike Guo, Yaodong Yang

Challenge:	Using large-scale annotation data, large language models can generate noise, errors and biases, leading to unexpected behaviours.
Approach:	They propose a dataset to promote safety alignment in large language models . they separate helpfulness and harmlessness annotations for question-answering pairs .
Outcome:	The proposed dataset provides 44.6k prompts and 265k question-answer pairs with safety meta-labels for 19 harm categories and three severity levels, with answers generated by Llama-family models.

ContextLens: Modeling Imperfect Privacy and Safety Context for Legal Compliance (2026.acl-long)

Copied to clipboard

Haoran Li, Yulin Chen, Huihao Jing, Wenbin Hu, Tsz Ho Li, Chanhou Lou, Hong Ting Tsang, Sirui Han, Yangqiu Song

SafeLawBench: Towards Safe Alignment of Large Language Models (2025.findings-acl)

Challenge:	Existing approaches to contextualize safety and privacy assessments assume the availability of complete and clear context, whereas real-world contexts tend to be ambiguous and incomplete.
Approach:	They propose a semi-rule-based framework that leverages large language models to ground the input context in the legal domain and explicitly identify both known and unknown factors for legal compliance.
Outcome:	The proposed framework can significantly improve existing baselines without training and can identify the ambiguous and missing factors.

Copied to clipboard

Chuxue Cao, Han Zhu, Jiaming Ji, Qichao Sun, Zhenghao Zhu, Wu Yinyu, Josef Dai, Yaodong Yang, Sirui Han, Yike Guo

When Slower Isn’t Truer: Inverse Scaling Law of Truthfulness in Multimodal Reasoning (2026.findings-acl)

Challenge:	Recent studies indicate that large language models (LLMs) may exhibit risks, including threats to the protection of private data and the generation of hallucinations.
Approach:	They propose to evaluate LLMs from a legal perspective using the SafeLawBench benchmark.
Outcome:	The proposed framework categorizes safety risks into three levels based on legal standards and includes 24,860 multi-choice questions and 1,106 open-domain question-answering tasks.

Copied to clipboard

Sitong Fang, Wenjing Cao, Jiahao Li, Xuyao Wang, Chi-Min Chan, Sirui Han, Juntao Dai, Yike Guo, Yaodong Yang, Jiaming Ji

Benchmarking Multi-National Value Alignment for Large Language Models (2025.findings-acl)

Challenge:	a study of slow reasoning models for multimodal reasoning finds that they are more prone to fabricating plausible yet false details when confronted with incomplete or misleading visual inputs.
Approach:	They conduct the first systematic study of the inverse scaling law in slow-thinking paradigms for multimodal reasoning.
Outcome:	The findings suggest that slower reasoning models are more prone to fabricating false details . the study analyzed 5,000-sample hierarchical prompt dataset by 50 participants .

Copied to clipboard

Chengyi Ju, Weijie Shi, Chengzhong Liu, Jiaming Ji, Jipeng Zhang, Ruiyuan Zhang, Jiajie Xu, Yaodong Yang, Sirui Han, Yike Guo

FinMME: Benchmark Dataset for Financial Multi-Modal Reasoning Evaluation (2025.acl-long)

Challenge:	Existing studies on large language models focus on ethical reviews, failing to capture the diversity of national values.
Approach:	They propose a national value extraction pipeline to efficiently construct value assessment datasets and a model-based model with instruction tagging to process raw data sources.
Outcome:	The proposed benchmark evaluates the alignment of LLMs with the values of five major nations: China, the United States, the UK, France, and Germany.

Copied to clipboard

Junyu Luo, Zhizhuo Kou, Liming Yang, Xiao Luo, Jinsheng Huang, Zhiping Xiao, Jingshu Peng, Chengzhong Liu, Jiaming Ji, Xuanzhe Liu, Sirui Han, Ming Zhang, Yike Guo

PrivaCI-Bench: Evaluating Privacy with Contextual Integrity and Legal Compliance (2025.acl-long)

Challenge:	Multimodal Large Language Models (MLLMs) have experienced rapid development in recent years, but there is a notable lack of effective and specialized multimodal evaluation datasets in the financial domain.
Approach:	They introduce FinMME, a multimodal large language model with 11,000 financial research samples and 20 annotators.
Outcome:	The proposed model performs better than state-of-the-art models, highlighting its challenging nature.

Copied to clipboard

Haoran Li, Wenbin Hu, Huihao Jing, Yulin Chen, Qi Hu, Sirui Han, Tianshu Chu, Peizhao Hu, Yangqiu Song