Research map / Reinforcement learning / Visual and representation RL
Continuously evolving rewards in an open-ended environment
Richard M. Bailey
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 continuous learning (fixed reward) to inference-only policy (ablation) | fixed baseline policy · Coins challenge (Exp 2) | population survival time: populations survive ~50% longer | Improved |
| add evolution to inference-only policy (ablation) | fixed baseline policy · Coins challenge (Exp 2b) | population survival | No clear effect |
| add evolution to learning with fixed reward (ablation) | learning agents · Coins challenge (Exp 2d) | population survival | No clear effect |
| add RULE reward updating to continuous learning (ablation) | learning agents · Coins challenge (Exp 2e) | population survival at high coin rate: survives coin rates beyond 250 coins/cycle | Improved |
| add evolution to RULE at slower coin increase (ablation) | RULE agents · Coins challenge (Exp 2f, slow rate) | population size | Improved |
| add evolution to RULE at faster coin increase (ablation) | RULE agents · Coins challenge (Exp 2f, fast rate) | population survival | Hurt |
| make coins beneficial (vitamins) under RULE (ablation) | RULE agents · Vitamins (Exp 3) | coin reward coefficient: reaches maximum value of 1 in ~8,000 s (weak) vs ~4,000 s (strong) | Improved |
| make coins harmful (poison) under RULE (ablation) | RULE agents · Poison (Exp 3) | coin reward coefficient: reduces to near zero | Hurt |
| raise constant coin rate from <=30 to >=50 coins/cycle (ablation) | baseline policy · Coins sensitivity (Exp 1) | population collapse: <=30 no impact; >=50 always collapsed | Hurt |
Closest papers by meaning
- SEEA-R1: Tree-Structured Reinforcement Fine-Tuning for Self-Evolving Embodied Agents (NeurIPS 2025 · similarity 0.83)
- Abstract and Explore: A Novel Behavioral Metric with Cyclic Dynamics in Reinforcement Learning (AAAI 2024 · similarity 0.83)
- REvolve: Reward Evolution with Large Language Models using Human Feedback (ICLR 2025 · similarity 0.83)
- Reinforcement Learning with Adaptive Reward Modeling for Expensive-to-Evaluate Systems (ICML 2025 · similarity 0.83)
- Inverse Reinforcement Learning with Switching Rewards and History Dependency for Characterizing Animal Behaviors (ICML 2025 · similarity 0.83)
- Inverting the Bellman Equation: From $Q$-Values to World Models (NeurIPS 2026 · similarity 0.83)
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.
