Mining Intrinsic Rewards from LLM Hidden States for Efficient Best-of-N Sampling
Jizhou Guo, Zhaomin Wu, Hanchen Yang, Philip S. Yu
Why Reward Models Are a Bottleneck
Key Insight
SWIFT: Simple Weighted Intrinsic Feedback
Results: Accuracy & Generalization
Efficiency
Why Does SWIFT Work?
Many Thanks!
Jizhou Guo’s homepage: https://aster2024.github.io/