TY  - RPRT
TI  - Mining Intrinsic Rewards from LLM Hidden States for Efficient Best-of-N Sampling
AU  - Jizhou Guo
AU  - Zhaomin Wu
AU  - Hanchen Yang
AU  - Philip S. Yu
PY  - 2026
DO  - 10.1145/3770854.3780302
UR  - https://arxiv.org/abs/2505.12225
ID  - 2505.12225
ER  -