Research map / Vision-language and multimodal / Robot action and navigation
ENACT: Evaluating Embodied Cognition with World Modeling of Egocentric Interaction
Qineng Wang, Wenlong Huang, Yu Zhou, Hang Yin, Tianwei Bao, Jianwen Lyu, Weiyu Liu, Ruohan Zhang, Jiajun Wu, Li Fei-Fei, Manling Li
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 |
|---|---|---|---|
| increase interaction horizon L from 3 to 10 | GPT-5 · ENACT forward | pairwise accuracy: 84.62 -> 46.93 | Hurt |
| increase interaction horizon L from 3 to 10 | GPT-5 · ENACT inverse | pairwise accuracy: 86.28 -> 55.33 | Hurt |
| switch task from forward to inverse world modeling | GPT-5 · ENACT L=10 | pairwise accuracy: 46.93 -> 55.33 | Improved |
| use real-world imagery instead of simulator (ablation) | InternVL3.5-241B-A28B · ENACT real-world | pairwise accuracy | No clear effect |
| change rendering realism (Realistic, Path Tracing, Ray Tracing Only) (ablation) | GPT-5 mini · ENACT simulation, L in {3,6,9} | pairwise accuracy | No clear effect |
| widen camera aperture to 60, 80 or fisheye vs 40 (ablation) | GPT-5 mini · ENACT simulation | pairwise accuracy | Hurt |
| change camera aperture from 40 to 30 (ablation) | GPT-5 mini · ENACT simulation | pairwise accuracy | No clear effect |
| raise camera height by 0.5 m (ablation) | GPT-5 mini · ENACT forward | pairwise accuracy: Delta=-0.13 | Hurt |
| raise camera height by 0.5 m (ablation) | GPT-5 mini · ENACT inverse | pairwise accuracy: Delta=-0.06 | No clear effect |
| lower camera height by 0.25 m (ablation) | GPT-5 mini · ENACT simulation | pairwise accuracy | No clear effect |
| change robot appearance (white, random, skin color) (ablation) | GPT-5 mini · ENACT simulation | pairwise accuracy | No clear effect |
| left vs right hand actions | VLMs (GPT-5 mini) · ENACT forward, hand predicates | mixing rate: 9.38% left->right vs 4.67% right->left | Mixed |
Closest papers by meaning
- EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents (ICML 2025 · similarity 0.90)
- WorldSimBench: Towards Video Generation Models as World Simulators (ICML 2025 · similarity 0.89)
- World–Value–Action Model: Implicit Planning for Vision–Language–Action Systems (NeurIPS 2026 · similarity 0.89)
- Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models (NeurIPS 2026 · similarity 0.89)
- SoMi-ToM: Evaluating Multi-Perspective Theory of Mind in Embodied Social Interactions (NeurIPS 2025 · similarity 0.88)
- Learning 3D Persistent Embodied World Models (NeurIPS 2025 · similarity 0.88)
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.
