5 papers
Theoretical Tensions in RLHF: Reconciling Empirical Success with Inconsistencies in Social Choice Theory
Jiancong Xiao, Zhekun Shi, Kaizhao Liu +2
Despite its empirical success, Reinforcement Learning from Human Feedback (RLHF) has been shown to violate almost all the fundamental axioms in social choice theory -- such as majo…
Fundamental Limits of Game-Theoretic LLM Alignment: Smith Consistency and Preference Matching
Zhekun Shi, Kaizhao Liu, Qi Long +2
Nash Learning from Human Feedback is a game-theoretic framework for aligning large language models (LLMs) with human preferences by modeling learning as a two-player zero-sum game.…
Restoring Calibration for Aligned Large Language Models: A Calibration-Aware Fine-Tuning Approach
Jiancong Xiao, Bojian Hou, Zhanliang Wang +4
One of the key technologies for the success of Large Language Models (LLMs) is preference alignment. However, a notable side effect of preference alignment is poor calibration: whi…
Statistical Impossibility and Possibility of Aligning LLMs with Human Preferences: From Condorcet Paradox to Nash Equilibrium
Kaizhao Liu, Qi Long, Zhekun Shi +2
Aligning large language models (LLMs) with diverse human preferences is critical for ensuring fairness and informed outcomes when deploying these models for decision-making. In thi…
Magnetic Preference Optimization: Achieving Last-iterate Convergence for Language Model Alignment
Mingzhi Wang, Chengdong Ma, Qizhi Chen +7
Self-play methods have demonstrated remarkable success in enhancing model capabilities across various domains. In the context of Reinforcement Learning from Human Feedback (RLHF),…