Research map / Reinforcement learning / Visual and representation RL
Regularity as Intrinsic Reward for Free Play
Cansu Sancaktar, Justus Piater, Georg Martius
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 RaIR regularity reward to CEE-US ensemble disagreement (lambda=0.1) (ablation) | CEE-US · Construction Singletower 3 | zero-shot success rate: 0.40 ± 0.12 -> 0.75 ± 0.07 | Improved |
| add RaIR regularity reward to CEE-US ensemble disagreement (lambda=0.1) (ablation) | CEE-US · Construction Pyramid 6 | zero-shot success rate: 0.02 ± 0.01 -> 0.18 ± 0.04 | Improved |
| add RaIR regularity reward to CEE-US ensemble disagreement (lambda=0.1) (ablation) | CEE-US · Construction Multitower 2+2 | zero-shot success rate: 0.52 ± 0.05 -> 0.77 ± 0.06 | Improved |
| add RaIR to CEE-US (ablation) | CEE-US · Construction Pick&Place 6 | zero-shot success rate: 0.90 ± 0.02 -> 0.90 ± 0.02 | No clear effect |
| add RaIR to CEE-US (ablation) | CEE-US · Construction Throw 4 | zero-shot success rate: 0.49 ± 0.05 -> 0.32 ± 0.02 | Hurt |
| add RaIR to CEE-US (ablation) | CEE-US · Construction Flip 4 | zero-shot success rate: 0.73 ± 0.1 -> 0.63 ± 0.08 | Hurt |
| remove ensemble disagreement term (pure RaIR, lambda=0) (ablation) | RaIR + CEE-US · Construction Pyramid 5 | zero-shot success rate: 0.49 ± 0.06 -> 0.25 ± 0.05 | Hurt |
| remove ensemble disagreement term (pure RaIR, lambda=0) (ablation) | RaIR + CEE-US · Construction Pick&Place 6 | zero-shot success rate: 0.90 ± 0.02 -> 0.74 ± 0.05 | Hurt |
| remove ensemble disagreement term (pure RaIR, lambda=0) (ablation) | RaIR + CEE-US · Construction Flip 4 | zero-shot success rate: 0.63 ± 0.08 -> 0.65 ± 0.1 | No clear effect |
| use relative phi instead of absolute relative phi (ablation) | RaIR + CEE-US · Construction assembly tasks | zero-shot success rate: 0.75 ± 0.07 -> 0.80 ± 0.07 | No clear effect |
| increase lambda to 1 (more weight on ensemble disagreement) (ablation) | RaIR + CEE-US · Construction free play interactions | interaction metrics | Hurt |
| use controller mode best instead of sum for free play (ablation) | CEE-US · Construction Singletower 3 | zero-shot success rate: 0.26 ± 0.11 -> 0.40 ± 0.12 | Improved |
Closest papers by meaning
- Action-Dependent Optimality-Preserving Reward Shaping (ICML 2025 · similarity 0.87)
- A Single Goal is All You Need: Skills and Exploration Emerge from Contrastive RL without Rewards, Demonstrations, or Subgoals (ICLR 2025 · similarity 0.86)
- Eureka: Human-Level Reward Design via Coding Large Language Models (ICLR 2024 · similarity 0.86)
- Motif: Intrinsic Motivation from Artificial Intelligence Feedback (ICLR 2024 · similarity 0.86)
- SEEA-R1: Tree-Structured Reinforcement Fine-Tuning for Self-Evolving Embodied Agents (NeurIPS 2025 · similarity 0.86)
- Efficient Language-instructed Skill Acquisition via Reward-Policy Co-Evolution (AAAI 2025 · similarity 0.86)
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.
