Research map / Reinforcement learning / Offline RL
Survival Instinct in Offline Reinforcement Learning
Anqi Li, Dipendra Misra, Andrey Kolobov, Ching-An Cheng
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 |
|---|---|---|---|
| Train with zero, random, or negative rewards instead of the true reward | ATAC and PSPI · D4RL hopper-random, hopper-medium, and walker2d-medium | normalized score | Improved |
| Train with wrong rewards | Offline RL algorithms · D4RL medium-replay hopper and walker2d | normalized score | Hurt |
| Train with wrong rewards | Offline RL algorithms · D4RL halfcheetah datasets | normalized score | Mixed |
| Train with wrong rewards | ATAC and PSPI · D4RL hopper and walker2d | episode length | Improved |
| Train with wrong rewards | CQL · D4RL walker2d datasets | normalized score | Hurt |
| Train with wrong rewards | All offline RL algorithms · D4RL walker2d-expert and halfcheetah-expert | normalized score | Improved |
| Train with wrong rewards | ATAC and PSPI · Meta-World goal-oriented tasks | success rate | Mixed |
| Train with wrong rewards | IQL · Meta-World goal-oriented tasks | success rate | Hurt |
| Train with wrong rewards | Decision Transformer · Meta-World goal-oriented tasks | success rate | No clear effect |
| Train with wrong rewards | Offline RL algorithms · D4RL halfcheetah-medium-expert | normalized score | Hurt |
| Filter out transitions with non-zero cost before offline RL training | ATAC and PSPI · SafetyGymnasium | normalized reward | Improved |
Closest papers by meaning
- Bayesian Design Principles for Offline-to-Online Reinforcement Learning (ICML 2024 · similarity 0.91)
- Simple Ingredients for Offline Reinforcement Learning (ICML 2024 · similarity 0.90)
- A Unified Principle of Pessimism for Offline Reinforcement Learning under Model Mismatch (NeurIPS 2024 · similarity 0.90)
- From Static Policies to Adaptive Priors in Offline Reinforcement Learning (NeurIPS 2026 · similarity 0.90)
- Forecasting in Offline Reinforcement Learning for Non-stationary Environments (NeurIPS 2025 · similarity 0.90)
- Online Optimization for Offline Safe Reinforcement Learning (NeurIPS 2025 · similarity 0.90)
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.
