Xiaopeng Yuan

About

Xiaopeng Yuan is a first-year PhD student at the UIUC School of Information Sciences, advised by Haohan Wang. He studies LLM reasoning, test-time training, and long-context learning to make language models more capable and reliable on complex tasks. He hopes to build AI tools that people find genuinely useful and that benefit a wide range of users.

Research interests

• LLM reasoning
• Test-time training
• Long-context learning
• Reasoning in multi-agent systems

Publications

Peer-reviewed

• KV-PRM: Efficient Process Reward Modeling via KV-Cache Transfer for Multi-Agent Test-Time Scaling Kuang P, Jin H, Han X, Wang Y, Yuan X, Yu Y, Xu K, Wang H. NeurIPS 2026
• Agent Primitives: Reusable Latent Building Blocks for Multi-Agent Systems Jin H, Kuang P, Yu Y, Yuan X, Wang H. ICML 2026 [arXiv]
• Query-efficient Domain Knowledge Stealing against Large Language Models Li Z, Yuan X, Shen B, Le K, Wang H, Zhou X, Gao S, Dong Y. AAAI 2026

Preprints

• Let the Library Speak: Self-Advertised Method Selection for Formal Proving Yuan X, Wang S, Wang Y, Jin H, Kuang P, Wang J, Yu L, Wang H. Preprint, 2026
• Objects Without Morphisms: What LLMs for Mathematics Do Not Represent Wang Y, Wang S, Yuan X, Wang H. Preprint, 2026
• Composing Task-specific Agent Harnesses at Test Time with Reusable Primitives Kuang P, Jin H, Wu D, Deng F, Yuan X, Wang J, Wang H. Preprint, 2026
• ANTMAN: Adaptive Need Tracking for Multi-Agent Navigation in Large Information Spaces Wang J, Jin H, Yuan X, Kuang P, Wang H. Preprint, 2026
• Beyond Asking: A Pipeline for Personalized Game Generation that Reads Players from Behavior Lu Y, Yuan X, Wang H. Preprint, 2026
• EASE-TTT: Evidence-Aligned Selective Test-Time Training for Long-Context Question Answering Yuan X, Wang Z, Wang S, Yang Z, Wang H, Dong Y. Preprint, 2026
• Closing the Loop on Latent Reasoning via Test-Time Reconstruction Yuan X, Jin H, Yu Y, Kuang P, Yu L, Dong Y, Wang H. Preprint, 2026
• Do Self-Evolving Agents Forget? Capability Degradation and Preservation in Lifelong LLM Agent Adaptation Yu Y, Yuan X, Jin H, Liu H, Yu Y, Wang H. Preprint, 2026
• Learning to Communicate: Toward End-to-End Optimization of Multi-Agent Language Systems Yu Y, Liu H, Jin H, Yuan X, Kuang P, Wang H. Preprint, 2026
• Tracing LLM Reasoning Processes with Strategic Games: A Framework for Planning, Revision, and Resource-Constrained Decision Making Yuan X, Zhang X, Xu K, Xu Y, Yu L, Wang J, Dong Y, Wang H. Preprint, 2025

Full list on Google Scholar. More coming soon.