Research map / Reinforcement learning / Policy gradient and actor-critic
Local Reinforcement Learning with Action-Conditioned Root Mean Squared Q-Functions
Zequan Wu, Mengye Ren
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 action conditioning at the input (ablation) | ARQ · MinAtar Breakout and SpaceInvaders | average return: ~85 vs. ~55 | Improved |
| add action conditioning at the input (ablation) | AD · MinAtar Breakout and SpaceInvaders | average return: slight improvement | Improved |
| use RMS goodness instead of mean (ablation) | ARQ · MinAtar Breakout | average return: 79.84 -> 87.84 | Improved |
| use RMS goodness instead of mean squared (ablation) | ARQ · MinAtar Breakout | average return: 82.10 -> 87.84 | Improved |
| use RMS goodness instead of variance (ablation) | ARQ · MinAtar SpaceInvaders | average return: 416.46 -> 544.99 | Improved |
| use RMS goodness instead of mean squared (ablation) | ARQ · MinAtar SpaceInvaders | average return: 434.88 -> 544.99 | Improved |
| use RMS goodness instead of mean (ablation) | ARQ · MinAtar SpaceInvaders | average return: 500.13 -> 544.99 | Improved |
| replace AD scalar output with any ARQ goodness function (ablation) | AD · MinAtar Breakout | average return: 67.40 -> 81.34 | Improved |
| replace AD with ARQ at 0.5x parameter scale (ablation) | AD · MinAtar Breakout | average return: 66.34 -> 68.12 | Improved |
| replace AD with ARQ at 1x parameter scale (ablation) | AD · MinAtar Breakout | average return: 64.20 -> 86.26 | Improved |
| replace AD with ARQ at 1.5x parameter scale (ablation) | AD · MinAtar Breakout | average return: 56.63 -> 70.40 | Improved |
| replace AD with ARQ at 2x parameter scale (ablation) | AD · MinAtar Breakout | average return: 59.79 -> 83.26 | Improved |
Closest papers by meaning
- Stochastic Q-learning for Large Discrete Action Spaces (ICML 2024 · similarity 0.87)
- Long-Horizon Q-Learning: Accurate Value Learning via n-Step Inequalities (NeurIPS 2026 · similarity 0.87)
- Learning from Suboptimal Data in Continuous Control via Auto-Regressive Soft Q-Network (ICML 2025 · similarity 0.86)
- Model-based Offline Reinforcement Learning with Lower Expectile Q-Learning (ICLR 2025 · similarity 0.86)
- Enhancing Value Function Estimation through First-Order State-Action Dynamics in Offline Reinforcement Learning (ICML 2024 · similarity 0.86)
- Coarse-to-fine Q-Network with Action Sequence for Data-Efficient Reinforcement Learning (NeurIPS 2025 · 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.
