AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Token Hidden Reward: Steering Exploration-Exploitation in Group Relative Deep Reinforcement Learning

Wenlong Deng, Yi Ren, Yushu Li, Boying Gong, Danica J. Sutherland, Xiaoxiao Li, Christos Thrampoulidis

ICLR 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards · 8 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train only on dominant (high |THR|) tokens (ablation)GRPO · math benchmarks, greedy total avgaccuracy (%): 42.7 -> 43.3No clear effect
amplify positive-THR tokens (p=0.1) (ablation)THR · math benchmarks greedy total avgaccuracy (%): 43.3 -> 46.8Improved
amplify negative-THR tokens (p<0) (ablation)THR · Pass@256 average, AIME/AMCPass@256 (%): 72.5 -> 76.7Improved
amplify negative-THR tokens (p<0) vs positive on greedy accuracy (ablation)THR · math benchmarks greedy total avgaccuracy (%): Qwen2.5-Math-7B: 46.8 (p>0) vs 44.0 (p<0); 1.5B: 36.3 vs 36.4Mixed
use THR (p<0) instead of Pass@K-mixed trainingQwen2.5-Math-1.5B · Pass@K averagePass@256 (%): 73.3 -> 76.7Improved
combine THR (p<0) directly with Pass@K-mixed (ablation)THR (p<0) · Pass@K averagePass@1 (%): 29.7 -> 27.3 (Qwen2.5-Math-7B)Hurt
combine THR (p<0) with static chi=0.2 Pass@K mix (ablation)THR (p<0) · Pass@K averagePass@1 (%): 29.7 -> 30.2 (Qwen2.5-Math-7B)Improved
use THR (p<0) instead of Cov-KL entropy regularizationQwen2.5-Math-1.5B · Pass@K averagePass@256 (%): 69.2 -> 76.7Improved
add THR (p=-0.1) to GSPOGSPO · Pass@K averagePass@256 (%): 69.5 -> 73.3Improved
THR (p<0) vs GRPO on 7B at large K on AIME24Qwen2.5-Math-7B · AIME 2024Pass@256 (%): 76.7 -> 73.3Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy