Publications
You can also find my articles on Google Scholar.
AI for Math and Physics
- From Research Goals to Scientific Closure: Evaluating Agents Through Evidence-Supported Physics Research · arXiv
- SciWalker: Synthesizing Scientific Coding Problems with Operator Graphs and Execution Feedback · arXiv
- ComBench: A Benchmark for Rigorous Proof Reasoning and Constructive Realization in Olympiad-Level Combinatorics · arXiv
- Achieving Gold-Medal-Level Olympiad Reasoning via Simple and Unified Scaling · Technical Report
- Teaching Thinking Models to Reason with Tools: A Full-Pipeline Recipe for Tool-Integrated Reasoning · arXiv
- Sci-Verifier: Scientific Verifier with Thinking · ICLR 2026
- P1-VL: Bridging Visual Perception and Scientific Reasoning in Physics Olympiads · Technical Report
- P1: Mastering Physics Olympiads with Reinforcement Learning · Technical Report
- HiPhO: How Far Are (M)LLMs from Humans in the Latest High School Physics Olympiad Benchmark? · ICML 2026
Post-Training
- Intern-Decision: Efficient Multimodal Decision Making with Autoregressive Masked Language Modeling · Technical Report
- TPPO: Tracking-Value PPO For Language Model Reasoning · arXiv
- Gains and Collapse in On-Policy Distillation: A Reinforcement Learning Perspective · arXiv
- Rethinking Critic Learning in PPO: Understanding and Mitigating Value Flattening · arXiv
- SimpleOPD: Simple Tokenizer-Agnostic On-Policy Distillation for Long-Context Reasoning · arXiv
- Draft-OPD: On-Policy Distillation for Speculative Draft Models · EMNLP 2026
- TEMPO: Scaling Test-time Training for Large Reasoning Models · arXiv
- Spotlight on token perception for multimodal reinforcement learning · ICLR 2026
- Think Longer to Explore Deeper: Learn to Explore In-Context via Length-Incentivized Reinforcement Learning · arXiv
Forgetting of LLMs
- An empirical study of catastrophic forgetting in large language models during continual fine-tuning · IEEE Transactions on Audio, Speech and Language Processing 2025
- RefuteBench 2.0 -- Agentic Benchmark for Dynamic Evaluation of LLM Responses to Refutation Instruction · IEEE/ACM Transactions on Audio, Speech, and Language Processing
- RefuteBench: Evaluating Refuting Instruction-Following for Large Language Models · Findings of ACL 2024
- Mitigating Catastrophic Forgetting in Task-Incremental Continual Learning with Adaptive Classification Criterion · arXiv
- Investigating Forgetting in Pre-Trained Representations Through Continual Learning · arXiv
Opinion Mining
- PerSphere: A Comprehensive Framework for Multi-Faceted Perspective Retrieval and Summarization · ACL 2025
- XAL: EXplainable Active Learning Makes Classifiers Better Low-resource Learners · NAACL 2024
- Enhancing Argument Structure Extraction with Efficient Leverage of Contextual Information · Findings of EMNLP 2023
- Improving (Dis)agreement Detection with Inductive Social Relation Information From Comment-Reply Interactions · WWW 2023
- Cooperative evolution with opinion formation in a complex social environment · Chaos: An Interdisciplinary Journal of Nonlinear Science 2022
- Mere Contrastive Learning for Cross-Domain Sentiment Analysis · COLING 2022
- Exploiting Sentiment and Common Sense for Zero-shot Stance Detection · COLING 2022
- Hegselmann-Krause Opinion Model on High-order Network · Chinese Control and Decision Conference (CCDC) 2022
- Adapted Deffuant–Weisbuch model with implicit and explicit opinions · Physica A: Statistical Mechanics and its Applications 2022
- Challenges for Open-domain Targeted Sentiment Analysis · arXiv
- Social bots and mass media manipulated public opinion through dual opinion climate · Chinese Physics B 2022
- Opinion formation with zealots on temporal network · Communications in Nonlinear Science and Numerical Simulation 2021
- Dynamic mechanism of social bots interfering with public opinion in network · Physica A: Statistical Mechanics and its Applications 2020
- Discrete opinion dynamics with social bots on signed network · Chinese Control Conference (CCC) 2020
- Consensus for expressed and private opinions under self-persuasion · IFAC-PapersOnLine 2020
Graph Neural Network
- Efficient prediction of peptide self-assembly through sequential and graphical encoding · Briefings in Bioinformatics 2023
- Are gradients on graph structure reliable in gray-box attacks? · CIKM 2022
- What Does the Gradient Tell When Attacking the Graph Structure · arXiv
- Surrogate representation learning with isometric mapping for gray-box graph adversarial attacks · WSDM 2022
- Towards reasonable budget allocation in untargeted graph structure attacks via gradient debias · NeurIPS 2022
Others
- DeepInstructor: An Agentic AI Instructor for Experience-Driven Idea Evaluation · arXiv
- Towards an AI Musician: Synthesizing Sheet Music Problems for Musical Reasoning · arXiv
- MCRanker: Generating Diverse Criteria On-the-Fly to Improve Pointwise LLM Rankers · WSDM 2025
- Keys to Robust Edits: from Theoretical Insights to Practical Advances · ACL 2025
- Task Calibration: Calibrating Large Language Models on Inference Tasks · Findings of ACL 2025
- SpearBot: Leveraging Large Language Models in a Generative-Critique Framework for Spear-Phishing Email Generation · Information Fusion
- Knowledge-Centric Hallucination Detection · EMNLP 2024
- OpenResearcher: Unleashing AI for Accelerated Scientific Research · EMNLP System Demonstrations 2024
- Scimine: An efficient systematic prioritization model based on richer semantic information · SIGIR 2023
- Game-Theoretic Control of Markov Chains: Two Applications · International Symposium on Mathematical Theory of Networks and Systems (MTNS) 2022
- Tactile–visual fusion based robotic grasp detection method with a reproducible sensor · International Journal of Computational Intelligence Systems 2021
- NS-DBSCAN: A density-based clustering algorithm in network space · ISPRS International Journal of Geo-Information 2019
- A Comparative Study of Two Strategies of Road Network Selection · Geomatics and Information Science of Wuhan University 2019