Research map / Video / Video-language understanding
StreamMind: Unlocking Full Frame Rate Streaming Video Dialogue through Event-Gated Cognition
Xin Ding, Hao Wu, Yifan Yang, Shiqi Jiang, Qianxi Zhang, Donglin Bai, Zhibo Chen, Ting Cao
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 balanced weight W_s to cross entropy (ablation) | Cognition Gate · Ego4D | TriggerAcc: 31.37% -> 43.34% | Improved |
| add balanced weight W_s to cross entropy (ablation) | Cognition Gate · SoccerNet | TriggerAcc: 31.34% -> 52.18% | Improved |
| use single LLM block gate instead of linear layer (ablation) | Cognition Gate · Ego4D | TriggerAcc: 20.13% -> 35.35% | Improved |
| use single LLM block gate instead of cross-attention + linear (ablation) | Cognition Gate · Ego4D | TriggerAcc: 24.34% -> 35.35% | Improved |
| initialize gate from early LLM blocks instead of random (ablation) | Cognition Gate · Ego4D | TriggerAcc: 39.65% -> 43.34% | Improved |
| initialize gate from early LLM blocks instead of SkipBlock (ablation) | Cognition Gate · Ego4D | TriggerAcc: 40.23% -> 43.34% | Improved |
| use 4 gate blocks instead of 2 (ablation) | Cognition Gate · Ego4D | TriggerAcc: 41.53% -> 43.34% | Improved |
| use 4 gate blocks instead of 6 (ablation) | Cognition Gate · Ego4D | TriggerAcc: 42.56% -> 43.34% | No clear effect |
| use EPFE instead of Q-former (ablation) | StreamMind · Ego4D | TriggerAcc: 26.65% -> 43.34% | Improved |
| use EPFE instead of STC connector (ablation) | StreamMind · Ego4D | TimeDiff: 3.56 -> 1.89 | Improved |
| raise balance weight from 0.15 to 0.20 (ablation) | Cognition Gate · Ego4D | TriggerAcc: 43.34% -> 35.35% | Hurt |
| raise balance weight from 0.03 to 0.05 (ablation) | Cognition Gate · SoccerNet | TriggerAcc: 52.18% -> 41.22% | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- VideoLLM-Online — VideoLLM-online: Online Video Large Language Model for Streaming Video (CVPR 2024) · better on 8 of 8 shared results
- LION-FS — LION-FS: Fast & Slow Video-Language Thinker as Online Video Assistant (CVPR 2025) · better on 2 of 2 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
- StreamForest: Efficient Online Video Understanding with Persistent Event Memory (NeurIPS 2025 · similarity 0.87)
- Streaming Video Understanding and Multi-round Interaction with Memory-enhanced Knowledge (ICLR 2025 · similarity 0.86)
- Dispider: Enabling Video LLMs with Active Real-Time Interaction via Disentangled Perception, Decision, and Reaction (CVPR 2025 · similarity 0.86)
- StreamBridge: Turning Your Offline Video Large Language Model into a Proactive Streaming Assistant (NeurIPS 2025 · similarity 0.86)
- StreamGaze: Gaze-Guided Temporal Reasoning and Proactive Understanding in Streaming Videos (NeurIPS 2026 · similarity 0.86)
- VideoLLM-online: Online Video Large Language Model for Streaming Video (CVPR 2024 · similarity 0.85)
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.
