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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add sample-then-forget unlearning to GRPO rollouts | GRPO · Qwen2.5-3B math average | accuracy: 21.0 -> 26.1 | Improved |
| add sample-then-forget unlearning to GRPO rollouts | GRPO · Llama-3.2-3B-Instruct math average | accuracy: 17.6 -> 23.4 | Improved |
| add sample-then-forget unlearning to GRPO rollouts | GRPO · Qwen3-8B-Base math average | accuracy: 34.7 -> 38.3 | Improved |
| add sample-then-forget unlearning to GRPO rollouts | GRPO · Qwen3-14B-Base math average | accuracy: 40.6 -> 44.7 | Improved |
| add sample-then-forget unlearning to GRPO rollouts | GRPO · AMC23 Qwen2.5-3B | accuracy: 30.3 -> 45.0 | Improved |
| raise sampling temperature | GRPO · Qwen2.5-3B math average | accuracy: 21.0 -> 21.5 | Improved |
| raise sampling temperature to 1.5 | GRPO · Qwen2.5-3B | accuracy | Hurt |
| increase entropy coefficient | GRPO · Qwen2.5-3B math average | accuracy: 21.0 -> 23.9 | Improved |
| use DAPO clip higher | GRPO · Qwen3-8B-Base math average | accuracy: 34.7 -> 32.7 | Hurt |
| increase number of rollouts | GRPO · Qwen2.5-3B math average | accuracy: 21.0 -> 23.2 | Improved |
| add EEPO on top of increased entropy (ablation) | GRPO + Increased Entropy · Qwen2.5-3B math average | accuracy: 23.9 -> 27.9 | Improved |
| add EEPO on top of clip higher (ablation) | GRPO + Clip High · Qwen2.5-3B math average | accuracy: 22.9 -> 26.6 | Improved |
Closest papers by meaning
- LLM-Explorer: A Plug-in Reinforcement Learning Policy Exploration Enhancement Driven by Large Language Models (NeurIPS 2025 · similarity 0.91)
- Exploratory Memory-Augmented LLM Agent via Hybrid On- and Off-Policy Optimization (ICLR 2026 · similarity 0.91)
- Learning to Explore with Parameter-Space Noise: A Deep Dive into Parameter-Space Noise for Reinforcement Learning with Verifiable Rewards (NeurIPS 2026 · similarity 0.90)
- One-Way Policy Optimization for Self-Evolving LLMs (ICML 2026 · similarity 0.90)
- Stratified GRPO: Handling Structural Heterogeneity in Reinforcement Learning of LLM Search Agents (ICML 2026 · similarity 0.89)
- Prompted Policy Search: Reinforcement Learning through Linguistic and Numerical Reasoning in LLMs (NeurIPS 2025 · similarity 0.89)
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.
