Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Learning to Play With Intrinsically-Motivated, Self-Aware Agents
Nick Haber, Damian Mrowca, Stephanie Wang, Li F Fei-Fei, Daniel L Yamins
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 |
|---|---|---|---|
| Use the self-model policy to seek high world-model prediction loss instead of a random policy (ablation) | ID-SP · Hard object interaction (one object) | f accuracy: 42.1 -> 51.1 | Improved |
| Use the self-model policy to seek high world-model prediction loss instead of a random policy (ablation) | ID-SP · Hard object interaction (one object) | tau accuracy: 41.3 -> 43.2 | Improved |
| Use an active policy rather than a random policy (ablation) | ID-SP · Object recognition transfer | Recognition accuracy: 12.3 -> 28.5 | Improved |
| Freeze the world-model encoding at random weights (ablation) | ID-SP · Hard object interaction (one object) | f accuracy: 51.1 -> 33.1 | Hurt |
| Freeze the world-model encoding at random weights (ablation) | ID-SP · Hard object interaction (one object) | tau accuracy: 43.2 -> 32.1 | Hurt |
| Use a latent-space future prediction world-model instead of inverse dynamics alone (ablation) | self-model-driven agent · Object localization transfer | Localization error [px]: 4.36 -> 5.94 | Hurt |
| Use a latent-space future prediction world-model instead of inverse dynamics alone (ablation) | self-model-driven agent · Object recognition transfer | Recognition accuracy: 28.5 -> 18.7 | Hurt |
| Train with two objects present instead of one (ablation) | ID-SP · Recognition transfer | Recognition accuracy: 28.5 -> 39.7 | Improved |
| Increase the ID-SP policy horizon to T=40 rather than use a shorter horizon such as T=5 (ablation) | ID-SP · Two-object interactions | two-object play state frequency | Improved |
| Use LF-SP rather than ID-SP with a long horizon for multi-object play (ablation) | self-model-driven agent · Two-object interactions | two-object play state frequency | Hurt |
Closest papers by meaning
- Behavioral Exploration: Learning to Explore via In-Context Adaptation (ICML 2025 · similarity 0.87)
- Intrinsic Goals for Autonomous Agents: Model-Based Exploration in Virtual Zebrafish Predicts Ethological Behavior and Whole-Brain Dynamics (NeurIPS 2025 · similarity 0.87)
- Heterogeneous Adversarial Play in Interactive Environments (NeurIPS 2025 · similarity 0.86)
- Wonder Wins Ways: Curiosity-Driven Exploration through Multi-Agent Contextual Calibration (NeurIPS 2025 · similarity 0.86)
- SENSEI: Semantic Exploration Guided by Foundation Models to Learn Versatile World Models (ICML 2025 · similarity 0.86)
- Motif: Intrinsic Motivation from Artificial Intelligence Feedback (ICLR 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.
