Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Intrinsically Motivated Goal Exploration Processes with Automatic Curriculum Learning
Sébastien Forestier, Rémy Portelas, Yoan Mollard, Pierre-Yves Oudeyer
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 modular (per-object) goal spaces instead of a flat goal space (ablation) | IMGEP agent · 2D Simulated, Magnet Tool | median exploration after 100k iterations: about 10% -> about 60% | Improved |
| add modular goal spaces instead of flat goal space (ablation) | IMGEP agent · Minecraft Mountain Cart, Pickaxe | median exploration: 15% -> 45% | Improved |
| add modular goal spaces (ablation) | IMGEP agent · Robotic environment | exploration of objects other than hand | Improved |
| add learning-progress-based goal space selection instead of random choice (AMB vs RMB) (ablation) | Modular IMGEP · Minecraft Mountain Cart, Pickaxe competence | median competence: 40 -> 45 | Improved |
| add learning-progress goal space selection instead of random (AMB vs RMB) (ablation) | Modular IMGEP · Minecraft Mountain Cart, Cart competence | median competence: 9 -> 11 | No clear effect |
| add learning-progress goal space selection (AMB vs RMB) (ablation) | Modular IMGEP · 2D Simulated, Magnet Tool | median exploration: 36 -> 61 | Improved |
| add learning-progress goal space selection (AMB vs RMB) (ablation) | Modular IMGEP · Minecraft Mountain Cart, Cart space | median unique outcomes: 162 -> 360 | Improved |
| replace hand-engineered fixed curriculum with autonomous learning-progress curriculum (AMB vs FC) (ablation) | Modular IMGEP · 2D Simulated, Magnet Tool | median exploration: 67 -> 61 | No clear effect |
| replace fixed curriculum with AMB (ablation) | Modular IMGEP · Minecraft Mountain Cart, Cart competence | median competence: 17 -> 11 | No clear effect |
| replace fixed curriculum with AMB (ablation) | Modular IMGEP · Robotic environment, right joystick and sound | exploration | Improved |
| use only a single goal space (SGS) instead of AMB (ablation) | Modular IMGEP · Minecraft Mountain Cart, Cart goals | median competence: 11 -> 0 | Hurt |
| use random policies instead of AMB (ablation) | IMGEP agent · 2D Simulated | touching toy with stick | Hurt |
Closest papers by meaning
- Intrinsic Goals for Autonomous Agents: Model-Based Exploration in Virtual Zebrafish Predicts Ethological Behavior and Whole-Brain Dynamics (NeurIPS 2025 · similarity 0.86)
- Behavioral Exploration: Learning to Explore via In-Context Adaptation (ICML 2025 · similarity 0.85)
- Efficient Language-instructed Skill Acquisition via Reward-Policy Co-Evolution (AAAI 2025 · similarity 0.85)
- A Single Goal is All You Need: Skills and Exploration Emerge from Contrastive RL without Rewards, Demonstrations, or Subgoals (ICLR 2025 · similarity 0.85)
- Exploring the Edges of Latent State Clusters for Goal-Conditioned Reinforcement Learning (NeurIPS 2024 · similarity 0.85)
- Latent Learning Progress Drives Autonomous Goal Selection in Human Reinforcement Learning (NeurIPS 2024 · similarity 0.84)
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.
