Research map / Reinforcement learning / Visual and representation RL
Reinforcement Learning-Guided Data Selection via Redundancy Assessment
Suorong Yang, Peijia Li, Furao Shen, Jian Zhao
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 the RL module (A2C with epsilon-cover reward) instead of a one-shot epsilon-coverage score from a pretrained model (ablation) | RL-Selector · CIFAR-10 / Tiny-ImageNet | test accuracy | Improved |
| use the transferable variant (15-epoch fine-tune of pretrained policy) | RL-Selector · Selection cost | selection speedup: more than 7.26x | Improved |
| use the transferable variant instead of full selection | RL-Selector · Selected-dataset accuracy | accuracy: slight decline; accelerates by nearly 7.14x | Mixed |
| select 90% of ImageNet-1k instead of training on the full set | RL-Selector · ImageNet-1k ViT-L | test accuracy: 84.50 -> 84.63 | Improved |
| select 90% of ImageNet-1k instead of training on the full set | RL-Selector · ImageNet-R (ResNet-50) | test accuracy: 36.16 -> 36.82 | Improved |
| select 90% of ImageNet-1k instead of training on the full set | RL-Selector · ImageNet-A (ResNet-50) | test accuracy: 3.10 -> 3.20 | Improved |
| select 90% of CIFAR-10 instead of training on the full set | RL-Selector · CIFAR-10 (ResNet-50) | test accuracy: 96.12 -> 96.38 | Improved |
| use RL-Selector instead of Moderate-DS at 20% selection on ViT-small | RL-Selector · CIFAR-10 (ViT-small) | test accuracy: 68.69 -> 69.21 | Improved |
| use RL-Selector instead of Forgetting at 40% selection | RL-Selector · CIFAR-10 (ResNet-50) | test accuracy: 93.61 -> 93.02 | Hurt |
| apply selection (clustering quality of selected data) | RL-Selector · CIFAR-10 features | Dunn Index: 92% higher | Improved |
Closest papers by meaning
- DUDS: Dual-stage Data Selection for Efficient Reinforcement Learning with Verifiable Rewards (NeurIPS 2026 · similarity 0.89)
- A Snapshot of Influence: A Local Data Attribution Framework for Online Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- UFO-RL: Uncertainty-Focused Optimization for Efficient Reinforcement Learning Data Selection (NeurIPS 2025 · similarity 0.87)
- The Impact of On-Policy Parallelized Data Collection on Deep Reinforcement Learning Networks (ICML 2025 · similarity 0.87)
- Understanding when Dynamics-Invariant Data Augmentations Benefit Model-free Reinforcement Learning Updates (ICLR 2024 · similarity 0.86)
- Enhancing Diversity In Parallel Agents: A Maximum State Entropy Exploration Story (ICML 2025 · similarity 0.86)
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.
