Research map / 3D vision / 3D generation
Embodied VideoAgent: Persistent Memory from Egocentric Videos and Embodied Sensors Enables Dynamic Scene Understanding
Yue Fan, Xiaojian Ma, Rongpeng Su, Jun Guo, Rujie Wu, Xi Chen, Qing 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 |
|---|---|---|---|
| use visual similarity for object re-ID instead of text only (ablation) | Embodied VideoAgent · Ego4D VQ3D val | Succ%: 53.05 -> 85.37 | Improved |
| use visual similarity for object re-ID instead of text only (ablation) | Embodied VideoAgent · Ego4D VQ3D val | QwP%: 56.10 -> 92.07 | Improved |
| use visual similarity for object re-ID instead of text only (ablation) | Embodied VideoAgent · Ego4D VQ3D val | Succ*%: 94.57 -> 92.72 | Hurt |
| use open-vocabulary detector with aggressive predictions vs EgoLoc | Embodied VideoAgent(image) · Ego4D VQ3D val | Succ%: 80.49 -> 85.37 | Improved |
| use open-vocabulary detector vs EgoLoc | Embodied VideoAgent(image) · Ego4D VQ3D val | L2: 1.45 -> 1.86 | Hurt |
| swap InternVL2-8B vqa tool for GPT-4o (ablation) | Embodied VideoAgent · OpenEQA subset | accuracy ALL: 41.2 -> 47.0 | Improved |
| use persistent object memory agent instead of VideoAgent | Embodied VideoAgent(InternVL2-8B) · OpenEQA subset | accuracy ALL: 36.3 -> 41.2 | Improved |
| use agentic system instead of end-to-end VLM | Embodied VideoAgent(GPT-4o) · OpenEQA subset | accuracy ALL: 30.6 -> 47.0 | Improved |
| replace GPT-4 w/ CG scene graph with frame captions (LLaVA-1.5) | GPT-4 · OpenEQA full | accuracy ALL: 36.5 -> 43.6 | Improved |
| use Embodied VideoAgent instead of VideoAgent | Embodied VideoAgent · EnvQA | Events accuracy: 5.54 -> 25.91 | Improved |
| use Embodied VideoAgent instead of VideoAgent | Embodied VideoAgent · EnvQA | States accuracy: 12.5 -> 35.50 | Improved |
| use Embodied VideoAgent instead of VideoAgent | Embodied VideoAgent · EnvQA | Orders accuracy: 65.5 -> 68.00 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- EgoLoc — EgoLoc: Revisiting 3D Object Localization from Egocentric Videos with Visual Queries (ICCV 2023) · better on 2 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- EmbodiedScan: A Holistic Multi-Modal 3D Perception Suite Towards Embodied AI (CVPR 2024 · similarity 0.90)
- EXPLORE-Bench: Egocentric Scene Prediction with Long-Horizon Reasoning (ECCV 2026 · similarity 0.89)
- 3D-Mem: 3D Scene Memory for Embodied Exploration and Reasoning (CVPR 2025 · similarity 0.88)
- Embodied Agents Meet Personalization: Investigating Challenges and Solutions Through the Lens of Memory Utilization (ICLR 2026 · similarity 0.88)
- ECBench: Can Multi-modal Foundation Models Understand the Egocentric World? A Holistic Embodied Cognition Benchmark (CVPR 2025 · similarity 0.87)
- SPA: 3D Spatial-Awareness Enables Effective Embodied Representation (ICLR 2025 · similarity 0.87)
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.
