Research map / Video / Video-language understanding
EgoSAT: A Comprehensive Benchmark of Egocentric Streaming Interaction Understanding
Yijia Lei, Jinzhao Li, Yichi Zhang, Jiacheng Hua, Yin Li, Miao Liu
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 |
|---|---|---|---|
| add hand and gaze ROI-prioritised KV token caching (ablation) | TimeChat-Online-7B · EgoSAT Now narration (MCQ) | accuracy: 33.33 -> 36.47 | Improved |
| add hand and gaze ROI-prioritised KV token caching (ablation) | TimeChat-Online-7B · EgoSAT Short-horizon retrieval (MCQ) | accuracy: 37.84 -> 34.41 | Hurt |
| add hand and gaze ROI-prioritised KV token caching (ablation) | TimeChat-Online-7B · EgoSAT Multi-step anticipation (MCQ) | average accuracy: 26.01 -> 25.05 | Hurt |
| add hand and gaze ROI-prioritised KV token caching (ablation) | TimeChat-Online-7B · EgoSAT Multi-step retrieval (MCQ) | average accuracy: 31.63 -> 30.82 | Hurt |
| add lightweight LoRA task SFT on EgoSAT | TimeChat-Online-7B · EgoSAT Short-horizon anticipation (MCQ) | accuracy: 29.72 -> 61.63 | Improved |
| add lightweight LoRA task SFT on EgoSAT | TimeChat-Online-7B · EgoSAT Now narration (MCQ) | accuracy: 33.33 -> 45.59 | Improved |
| add lightweight LoRA task SFT on EgoSAT | ROI-TimeChat-Online · EgoSAT Multi-step retrieval (MCQ) | average accuracy: 30.82 -> 41.24 | Improved |
| add LoRA SFT with MCQ-only schema | TimeChat-Online-7B · EgoSAT Now state switch | Fg->bg accuracy: 0.00 -> 0.00 | No clear effect |
| remove visual input (blind text-only model) (ablation) | Gemini 2.5 Pro · EgoSAT Short-horizon anticipation (MCQ) | accuracy: 19.37 -> 29.92 | Improved |
| remove visual input (blind text-only model) (ablation) | Qwen2.5-VL-32B · EgoSAT Short-horizon retrieval (MCQ) | accuracy: 31.58 -> 50.86 | Improved |
| remove visual input (blind text-only model) (ablation) | TimeChat-Online-7B · EgoSAT Short-horizon anticipation (MCQ) | accuracy: 18.01 -> 29.72 | Improved |
| use offline Qwen2.5-VL-72B instead of online TimeChat-Online-7B | video LLM · EgoSAT Multi-step retrieval (MCQ) | average accuracy: 31.63 -> 43.46 | Improved |
Closest papers by meaning
- Omnia de EgoTempo: Benchmarking Temporal Understanding of Multi-Modal LLMs in Egocentric Videos (CVPR 2025 · similarity 0.91)
- EgoThinker: Unveiling Egocentric Reasoning with Spatio-Temporal CoT (NeurIPS 2025 · similarity 0.91)
- EgoExoBench: A Benchmark for First- and Third-person View Video Understanding in MLLMs (NeurIPS 2025 · similarity 0.90)
- StreamGaze: Gaze-Guided Temporal Reasoning and Proactive Understanding in Streaming Videos (NeurIPS 2026 · similarity 0.90)
- EARL: Towards a Unified Analysis-Guided Reinforcement Learning Framework for Egocentric Interaction Reasoning and Pixel Grounding (ICML 2026 · similarity 0.88)
- In the Eye of MLLM: Benchmarking Egocentric Video Intent Understanding with Gaze-Guided Prompting (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.
