Research map / Reinforcement learning / Offline RL
Anti-Exploration by Random Network Distillation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use FiLM conditioning in the RND prior instead of concatenation (ablation) | SAC-RND actor bonus minimization · D4RL medium datasets: HalfCheetah, Walker2d, and Hopper | actor's ability to minimize the anti-exploration bonus | Helped |
| Use concatenation conditioning in the RND prior (ablation) | SAC-RND actor bonus minimization · D4RL medium datasets: HalfCheetah, Walker2d, and Hopper | actor's ability to minimize the anti-exploration bonus | Hurt |
| Use bilinear rather than concatenation conditioning in the predictor, with the prior fixed to FiLM (ablation) | SAC-RND · D4RL Gym v2 conditioning study (Average) | normalized score: 89.6 -> 99.9 | Helped |
| Use gating rather than concatenation conditioning in the predictor, with the prior fixed to FiLM (ablation) | SAC-RND · D4RL Gym v2 conditioning study (Average) | normalized score: 89.6 -> 67.3 | Hurt |
| Use FiLM rather than concatenation conditioning in the predictor, with the prior fixed to FiLM (ablation) | SAC-RND · D4RL Gym v2 conditioning study (Average) | normalized score: 89.6 -> 95.0 | Helped |
| Use predictor conditioning on the last layer rather than other tested depths (ablation) | SAC-RND actor cloning · D4RL medium datasets: HalfCheetah, Walker2d, and Hopper | mean squared error to behavioral-policy actions | Helped |
| Use prior conditioning on all layers rather than conditioning only at a single layer (ablation) | SAC-RND actor cloning · D4RL medium datasets: HalfCheetah, Walker2d, and Hopper | mean squared error to behavioral-policy actions | Helped |
| Use all-layer bilinear conditioning instead of FiLM conditioning in the prior (ablation) | SAC-RND · D4RL medium datasets: HalfCheetah, Walker2d, and Hopper | actor cloning performance | No clear effect |
Closest papers by meaning
- Exploration and Anti-Exploration with Distributional Random Network Distillation (ICML 2024 · similarity 0.90)
- Offline Behavior Distillation (NeurIPS 2024 · similarity 0.86)
- Selective Uncertainty Propagation in Offline RL (AAAI 2025 · similarity 0.86)
- OCEAN-MBRL: Offline Conservative Exploration for Model-Based Offline Reinforcement Learning (AAAI 2024 · similarity 0.86)
- Information-Directed Offline-to-Online Reinforcement Learning (NeurIPS 2026 · similarity 0.86)
- Doubly Mild Generalization 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.