Research map / Reinforcement learning / RL from feedback
Dynamic Important Example Mining for Reinforcement Finetuning
Haoru Tan, Sitong Wu, Yanfeng Chen, Shizhen Zhao, Yang-Tian Sun, Tianjia Liu, Chirui Chang, Shaofeng Zhang, Samm Sun, Xiuzhe Wu, Ruobing Xie, Xiaojuan Qi
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 dynamic gradient-alignment influence score instead of random value (ablation) | DIEM · MathVerse | accuracy: 53.0 -> 58.0 | Improved |
| add dynamic influence score instead of Pass@k score (ablation) | DIEM · MathVerse | accuracy: 53.2 -> 58.0 | Improved |
| add dynamic influence score instead of Pass@k distance to median (ablation) | DIEM · MathVerse | accuracy: 54.9 -> 58.0 | Improved |
| add dynamic influence score instead of PCL difficulty score (ablation) | DIEM · MathVerse | accuracy: 52.1 -> 58.0 | Improved |
| add dynamic influence score instead of difficulty distance to median (ablation) | DIEM · MathVerse | accuracy: 53.6 -> 58.0 | Improved |
| add constrained reweighting step (ablation) | DIEM · MathVerse | accuracy: 55.4 -> 58.0 | Improved |
| add constrained reweighting instead of softmax normalization (ablation) | DIEM · MathVerse | accuracy: 56.4 -> 58.0 | Improved |
| add DIEM to GRPO | GRPO · Math reasoning avg, Qwen3-4B | accuracy: 37.30 -> 40.66 | Improved |
| add DIEM to GRPO | GRPO · Math reasoning avg, Qwen2.5-7B | accuracy: 34.00 -> 35.68 | Improved |
| add DIEM to GRPO | GRPO · AIME25, Qwen2.5-7B | accuracy: 5.5 -> 10.8 | Improved |
| add DIEM to GRPO | GRPO · AMC-23, Qwen3-4B | accuracy: 58.5 -> 55.0 | Hurt |
| add DIEM to Vanilla RFT | Vanilla RFT · Multimodal avg, Qwen2.5-VL-7B | accuracy: 59.1 -> 61.8 | Improved |
Closest papers by meaning
- Dynamics-Predictive Sampling for Active RL Finetuning of Large Reasoning Models (ICLR 2026 · similarity 0.85)
- LICORICE: Label-Efficient Concept-Based Interpretable Reinforcement Learning (ICLR 2025 · similarity 0.85)
- Iteratively Refined Behavior Regularization for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.85)
- BOTS: A Unified Framework for Bayesian Online Task Selection in LLM Reinforcement Finetuning (ICLR 2026 · similarity 0.85)
- In-Context Reinforcement Learning From Suboptimal Historical Data (ICML 2025 · similarity 0.85)
- On Predictability of Reinforcement Learning Dynamics for Large Language Models (ICLR 2026 · similarity 0.85)
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.
