Research map / Reinforcement learning / Policy gradient and actor-critic
DisCor: Corrective Feedback in Reinforcement Learning via Distribution Correction
Aviral Kumar, Abhishek Gupta, Sergey Levine
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 |
|---|---|---|---|
| Reweight SAC replay-buffer transitions with DisCor | SAC · MT10 | success rate: 50% | Improved |
| Reweight SAC replay-buffer transitions with DisCor | SAC · MT10 | number of tasks making progress: 7/10 vs 3/10 | Improved |
| Reweight transitions using estimated target-value error rather than unweighted replay sampling | SAC · Meta-World robotic manipulation tasks | success rate | Improved |
| Reweight transitions using estimated target-value error rather than unweighted replay sampling | SAC · OpenAI Gym benchmarks with stochastic rewards | sample efficiency | Improved |
| Reweight replay-buffer transitions with DisCor | DQN · Atari: Pong, Breakout, Asterix | learning speed and performance | Improved |
| Prioritize transitions by Bellman error instead of using DisCor weights | SAC · Meta-World robotic manipulation tasks | performance | Hurt |
| Use the true value error instead of the estimated error bound Δ (ablation) | DisCor · Tabular MDPs | value error | Improved |
| Use estimated error bound Δ instead of the true value error (ablation) | DisCor · Tabular MDPs | performance | Improved |
| Prioritize transitions by high Bellman error rather than using DisCor's estimated target-error weighting (ablation) | DisCor · Tabular MDPs with sparse rewards | performance | Hurt |
Closest papers by meaning
- Long-Horizon Q-Learning: Accurate Value Learning via n-Step Inequalities (NeurIPS 2026 · similarity 0.88)
- Adaptive $Q$-Network: On-the-fly Target Selection for Deep Reinforcement Learning (ICLR 2025 · similarity 0.87)
- Reinforcement Learning from Imperfect Corrective Actions and Proxy Rewards (ICLR 2025 · similarity 0.87)
- $Q\sharp$: Provably Optimal Distributional RL for LLM Post-Training (NeurIPS 2025 · similarity 0.87)
- Diffusion Policies Creating a Trust Region for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- Exclusively Penalized Q-learning for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
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.
