Research map / Video / Video-language understanding
EgoEverything: A Benchmark for Human Behavior–Inspired Long-Context Egocentric Video Understanding in AR Environment
Qiance Tang, Ziqi Wang, Jieyu Lin, Ziyun Li, Barbara Salvo, Sai Qian Zhang
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 |
|---|---|---|---|
| restrict input to gaze crop (GC) instead of full frame (ablation) | Videollama3-7b · EgoEverything MCQ | accuracy: 49.1 -> 42.4 | Hurt |
| mask out gaze region (GM) instead of full frame (ablation) | Videollama3-7b · EgoEverything MCQ | accuracy: 49.1 -> 35.2 | Hurt |
| average-downsample frames to 10% resolution instead of full resolution (ablation) | Gemini 1.5 pro · EgoEverything MCQ | accuracy: 63.1 -> 58.4 | Hurt |
| use average downsampling instead of gaze crop (ablation) | Gemini 1.5 pro · EgoEverything MCQ | accuracy: 52.7 -> 58.4 | Improved |
| use gaze crop instead of gaze mask (ablation) | Gemini 1.5 pro · EgoEverything MCQ | accuracy: 37.7 -> 52.7 | Improved |
| add VideoMindPalace scene-graph pipeline over base VLM | Gemini 1.5 pro · EgoEverything MCQ | accuracy: 63.1 -> 33.2 | Hurt |
| add AMEGO scene-graph pipeline over base VLM | Videollama3-7b · EgoEverything MCQ | accuracy: 49.1 -> 19.5 | Hurt |
| add saliency-map gaze prompting | Videollama3-7b · EgoEverything MCQ | accuracy: 49.1 -> 51.19 | Improved |
| increase recall interval from smallest 10% to largest 10% (ablation) | Videollama3-7b · EgoEverything MCQ | accuracy: 49 -> 33.3 | Hurt |
| move target object from 100-200 px to 600-700 px from gaze (ablation) | Videollama3-7b · EgoEverything MCQ | accuracy: 54.8 -> 31.4 | Hurt |
| increase target object bounding-box area threshold (ablation) | evaluated VLMs · EgoEverything MCQ | accuracy | Improved |
| remove video frames (text-only input) (ablation) | Videollama3-7b, Gemini, LongVA · EgoEverything MCQ | accuracy | Hurt |
Closest papers by meaning
- In the Eye of MLLM: Benchmarking Egocentric Video Intent Understanding with Gaze-Guided Prompting (NeurIPS 2025 · similarity 0.90)
- Omnia de EgoTempo: Benchmarking Temporal Understanding of Multi-Modal LLMs in Egocentric Videos (CVPR 2025 · similarity 0.89)
- MMEgo: Towards Building Egocentric Multimodal LLMs for Video QA (ICLR 2025 · similarity 0.89)
- Action Scene Graphs for Long-Form Understanding of Egocentric Videos (CVPR 2024 · similarity 0.89)
- StreamGaze: Gaze-Guided Temporal Reasoning and Proactive Understanding in Streaming Videos (NeurIPS 2026 · similarity 0.89)
- ANNEXE: Unified Analyzing, Answering, and Pixel Grounding for Egocentric Interaction (CVPR 2025 · similarity 0.89)
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.
