AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

EEPO: Exploration-Enhanced Policy Optimization via Sample-Then-Forget

Liang Chen, Xueting Han, Qizhou Wang, Bo Han, Jing Bai, Hinrich Schuetze, Kam-Fai Wong

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
add sample-then-forget unlearning to GRPO rolloutsGRPO · Qwen2.5-3B math averageaccuracy: 21.0 -> 26.1Improved
add sample-then-forget unlearning to GRPO rolloutsGRPO · Llama-3.2-3B-Instruct math averageaccuracy: 17.6 -> 23.4Improved
add sample-then-forget unlearning to GRPO rolloutsGRPO · Qwen3-8B-Base math averageaccuracy: 34.7 -> 38.3Improved
add sample-then-forget unlearning to GRPO rolloutsGRPO · Qwen3-14B-Base math averageaccuracy: 40.6 -> 44.7Improved
add sample-then-forget unlearning to GRPO rolloutsGRPO · AMC23 Qwen2.5-3Baccuracy: 30.3 -> 45.0Improved
raise sampling temperatureGRPO · Qwen2.5-3B math averageaccuracy: 21.0 -> 21.5Improved
raise sampling temperature to 1.5GRPO · Qwen2.5-3BaccuracyHurt
increase entropy coefficientGRPO · Qwen2.5-3B math averageaccuracy: 21.0 -> 23.9Improved
use DAPO clip higherGRPO · Qwen3-8B-Base math averageaccuracy: 34.7 -> 32.7Hurt
increase number of rolloutsGRPO · Qwen2.5-3B math averageaccuracy: 21.0 -> 23.2Improved
add EEPO on top of increased entropy (ablation)GRPO + Increased Entropy · Qwen2.5-3B math averageaccuracy: 23.9 -> 27.9Improved
add EEPO on top of clip higher (ablation)GRPO + Clip High · Qwen2.5-3B math averageaccuracy: 22.9 -> 26.6Improved

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