Publications
Incentivizing Truthful Language Models via Peer Elicitation Games
NeurIPS 2025, September 2025
This paper introduces Peer Elicitation Games (PEG), a training-free, game-theoretic framework for aligning LLMs.
Competitive Multi-Agent Delegation For LLM Reasoning
Submitted to ICML 2025
This paper introduces COMMAND, a competitive multi-agent delegation framework that treats LLM reasoning as a principal-agent delegation process, using competition and alignment incentives to elicit higher-quality answers.
LightAgent: Production-level Open-source Agentic Al Framework
arXiv preprint, September 2025
This paper propose LightAgent, a lightweight, open-source agentic framework that balances flexibility and simplicity by integrating core functionalities like Memory, Tools, and Tree of Thought.