Research map / Reinforcement learning / Offline RL
SPQR: Controlling Q-ensemble Independence with Spiked Random Model for Reinforcement Learning
Dohyeok Lee, Seungyub Han, Taehyun Cho, Jungwoo Lee
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 SPQR independence regularization to SAC-Min | SAC-Min · D4RL Gym average | normalized average return: 84.5 -> 89.6 | Improved |
| Add SPQR independence regularization to CQL-Min | CQL-Min · Franka Kitchen average | normalized average return: 48.2 -> 52.6 | Improved |
| Add SPQR independence regularization to CQL-Min | CQL-Min · Antmaze average | normalized average return: 73.1 -> 84.4 | Improved |
| Add SPQR regularization with β=0.1 to SAC-Min | SAC-Min · hopper-random (evaluated on hopper-full-replay) | performance: 31.3±0.0 -> 35.6±1.4 | Improved |
| Increase SPQR regularization gain from β=0.1 to β=100.0 (ablation) | SPQR-SAC-Min · hopper-random (evaluated on hopper-full-replay) | spike count: 2182 -> 698 | Improved |
| Increase SPQR regularization gain from β=0.1 to β=100.0 (ablation) | SPQR-SAC-Min · hopper-random (evaluated on hopper-full-replay) | independence test acceptance ratio: 80.4% -> 97.6% | Improved |
| Increase SPQR regularization gain from β=0.1 to β=100.0 (ablation) | SPQR-SAC-Min · hopper-random (evaluated on hopper-full-replay) | performance: 35.6±1.4 -> 32.9±0.8 | Hurt |
| Use β=0.01 instead of β=0.1 (ablation) | SPQR-SAC-Min · hopper-random (evaluated on hopper-full-replay) | performance: 35.6±1.4 -> 33.6±1.8 | Hurt |
| Use β=1.0 instead of β=0.1 (ablation) | SPQR-SAC-Min · hopper-random (evaluated on hopper-full-replay) | performance: 35.6±1.4 -> 32.6±0.4 | Hurt |
| Use β=10.0 instead of β=0.1 (ablation) | SPQR-SAC-Min · hopper-random (evaluated on hopper-full-replay) | performance: 35.6±1.4 -> 32.7±0.4 | Hurt |
| Add SPQR independence regularization to EDAC | EDAC · D4RL Gym tasks | normalized average return | Mixed |
Closest papers by meaning
- Ensemble Bootstrapping for Q-Learning (ICML 2021 · similarity 0.89)
- Entropy-regularized Diffusion Policy with Q-Ensembles for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- Regularized Q-Learning (NeurIPS 2024 · similarity 0.86)
- Adaptive $Q$-Network: On-the-fly Target Selection for Deep Reinforcement Learning (ICLR 2025 · similarity 0.85)
- Q-Distribution guided Q-learning for offline reinforcement learning: Uncertainty penalized Q-value via consistency model (NeurIPS 2024 · similarity 0.85)
- Exclusively Penalized Q-learning for Offline Reinforcement Learning (NeurIPS 2024 · 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.
