Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Synthesizing Environment-Aware Activities via Activity Sketches
Yuan-Hong Liao, Xavier Puig, Marko Boben, Antonio Torralba, Sanja Fidler
Check 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 residual updates to represent environment-state changes induced by generated actions (ablation) | Graph program generator · VirtualHome-Env test, ground-truth sketches and graphs | F1: 0.4 -> 0.436 | Improved |
| Add residual updates to represent environment-state changes induced by generated actions (ablation) | Graph program generator · VirtualHome-Env test, ground-truth sketches and graphs | Executability: 46.79% -> 51.1% | Improved |
| Add graph propagation to incorporate neighboring object relations (ablation) | Program generator · VirtualHome-Env test, ground-truth sketches and graphs | LCS: 0.39 -> 0.526 | Improved |
| Use three rather than two graph propagation steps (ablation) | ResActGraph · VirtualHome-Env test, ground-truth sketches and graphs | LCS: 0.519 -> 0.536 | Improved |
| Use three rather than two graph propagation steps (ablation) | ResActGraph · VirtualHome-Env test, ground-truth sketches and graphs | F1: 0.436 -> 0.419 | Hurt |
| Replace residual graph-state updates with a GRU (ablation) | Graph program generator · VirtualHome-Env test, ground-truth sketches and graphs | F1: 0.436 -> 0.429 | Hurt |
| Replace residual graph-state updates with a fully connected layer (ablation) | Graph program generator · VirtualHome-Env test, ground-truth sketches and graphs | F1: 0.436 -> 0.4 | Hurt |
Closest papers by meaning
- Exploration-Driven Generative Interactive Environments (CVPR 2025 · similarity 0.84)
- ScaleEnv: Scaling Environment Synthesis from Scratch for Generalist Interactive Tool-Use Agent Training (ICML 2026 · similarity 0.83)
- WorldCoder, a Model-Based LLM Agent: Building World Models by Writing Code and Interacting with the Environment (NeurIPS 2024 · similarity 0.83)
- AgentStudio: A Toolkit for Building General Virtual Agents (ICLR 2025 · similarity 0.83)
- SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning (NeurIPS 2026 · similarity 0.83)
- Autonomous Continual Learning for Environment Adaptation of Computer-Use Agents (NeurIPS 2026 · similarity 0.82)
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.
