Research map / Video / Video-language understanding
Episodic Memory Question Answering
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 temporal observation features to the episodic scene memory (ablation) | EMQA · MP3D test, full tours | IoU: 27.42 { $\pm$ 0.64} -> 29.11 { $\pm$ 0.44} | Helped |
| Add temporal observation features to the episodic scene memory (ablation) | EMQA · Spatio-temporal localization questions | IoU: 24% relative improvement | Helped |
| Add temporal observation features to the episodic scene memory (ablation) | EMQA · Spatial localization questions | IoU | No clear effect |
| Use geometrically projected allocentric scene memory and LingUNet question grounding instead of direct top-down decoding | EMQA · MP3D test, full tours | IoU: 26.92 -> 27.42 | Helped |
| Use geometrically projected allocentric scene memory and LingUNet question grounding instead of direct top-down decoding | EMQA · MP3D test, full tours | Recall: 43.86 -> 60.81 | Helped |
| Replace spatially explicit scene memory with a language-only input | EMQA · MP3D test, full tours | IoU: 4.75 vs 29.11 | Hurt |
| Represent the tour with a compressed 1-D buffer of egocentric features | EMQA · MP3D test, full tours | IoU: 0.07; 0.01; 0.12 vs 29.11 | Hurt |
| Evaluate an oracle-pose-trained model with noisy pose inputs | EMQA · Noisy-pose evaluation | IoU: 39% drop | Hurt |
| Retrain SMNet and LingUNet using noisy-pose data | EMQA · Noisy-pose evaluation | IoU: 10% increase | Helped |
| Retrain the SMNet scene encoder using noisy-pose data | SMNetDecoder · Noisy-pose evaluation | IoU: 19% increase | Helped |
| Retrain SMNet using noisy-pose data | SMNet semantic map prediction · Noisy-pose evaluation | IoU: 6% gain | Helped |
| Use the proposed method rather than the strongest baseline under pose noise | EMQA · Noisy-pose evaluation | IoU drop: 29% vs 36% | Helped |
Closest papers by meaning
- WearVQA: A Visual Question Answering Benchmark for Wearables in Egocentric Authentic Real-world scenarios (NeurIPS 2025 · similarity 0.89)
- OpenEQA: Embodied Question Answering in the Era of Foundation Models (CVPR 2024 · similarity 0.87)
- Towards Comprehensive Scene Understanding: Integrating First and Third-Person Views for LVLMs (NeurIPS 2025 · similarity 0.87)
- Grounded Question-Answering in Long Egocentric Videos (CVPR 2024 · similarity 0.86)
- Omnia de EgoTempo: Benchmarking Temporal Understanding of Multi-Modal LLMs in Egocentric Videos (CVPR 2025 · similarity 0.86)
- EgoThink: Evaluating First-Person Perspective Thinking Capability of Vision-Language Models (CVPR 2024 · similarity 0.86)
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.