Research map / Label-efficient and robust learning / Recognition and long-tail
Behavior Is Everything: Towards Representing Concepts with Sensorimotor Contingencies
Nicholas Hay, Michael Stark, Alexander Schlegel, Carter Wendelken, Dennis Park, Eric Purdy, Tom Silver, D. Scott Phoenix, Dileep George
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 hierarchical SMCs with hand-designed curricula instead of flat policy | SMC policy · PixelWorld classification (avg) | accuracy: 74.5% -> 84.9% | Improved |
| train hierarchical SMCs with canonical curriculum instead of flat policy | SMC policy · PixelWorld classification (avg) | accuracy: 74.5% -> 80.2% | Improved |
| hierarchical SMCs with hand-designed curriculum instead of flat policy | SMC policy · PixelWorld bring-about (avg) | success rate: 52.7% -> 74.5% | Improved |
| hierarchical SMCs with canonical curriculum instead of flat policy | SMC policy · PixelWorld bring-about (avg) | success rate: 52.7% -> 75.0% | Improved |
| curriculum on object-number classification task | SMC policy · cl_self-left-of-two-objects_vs_one | accuracy: 52.6% -> 90.4% | Improved |
| curriculum on containment bring-about task | SMC policy · ba_self-in-container_from_between-potential-containers | success rate: 48.4% -> 82.0% | Improved |
| exclude primitive actions from hierarchical SMCs (curriculum) (ablation) | smc-curr · PixelWorld classification (avg) | accuracy: 84.9% -> 82.7% | No clear effect |
| exclude primitive actions from hierarchical SMCs (canonical) (ablation) | smc-canon · PixelWorld classification (avg) | accuracy: 80.3% -> 81.1% | No clear effect |
| exclude primitive actions from hierarchical SMCs (ablation) | smc-curr · PixelWorld bring-about (avg) | success rate: 74.5% -> 71.6% | No clear effect |
| exclude primitive actions from hierarchical SMCs (ablation) | smc-canon · PixelWorld bring-about (avg) | success rate: 75.1% -> 75.1% | No clear effect |
| interactive SMC vs passive CNN with global view | smc-base · PixelWorld classification (avg) | accuracy: 79.9% (cnn) vs 74.5% (smc-base) | Mixed |
Closest papers by meaning
- Behavioral Exploration: Learning to Explore via In-Context Adaptation (ICML 2025 · similarity 0.87)
- NeSyC: A Neuro-symbolic Continual Learner For Complex Embodied Tasks In Open Domains (ICLR 2025 · similarity 0.87)
- Ground-Compose-Reinforce: Grounding Language in Agentic Behaviours using Limited Data (NeurIPS 2025 · similarity 0.86)
- SENSEI: Semantic Exploration Guided by Foundation Models to Learn Versatile World Models (ICML 2025 · similarity 0.86)
- Embodied Agent Interface: Benchmarking LLMs for Embodied Decision Making (NeurIPS 2024 · similarity 0.85)
- Mental-Models for Multi-Agent Systems (NeurIPS 2026 · 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.
