Research map / Video / Video-language understanding
Em-Garde: A Propose-Match Framework for Proactive Streaming Video Understanding
Yikai Zheng, Xin Ding, Yifan Yang, Shiqi Jiang, Hao Wu, Qianxi Zhang, Weijun Wang, Ting Cao, Yunxin 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 IGPP proposal parser (ablation) | Em-Garde · OVO-Bench Forward Active Responding | F1: 24.0 -> 31.0 | Improved |
| add IGPP proposal parser (ablation) | Em-Garde · StreamingBench Proactive Output | accuracy: 28.8 -> 38.0 | Improved |
| add LPMM embedding matching (ablation) | Em-Garde · OVO-Bench Forward Active Responding | F1: 14.6 -> 31.0 | Improved |
| add LPMM embedding matching (ablation) | Em-Garde · StreamingBench Proactive Output | accuracy: 23.2 -> 38.0 | Improved |
| propose-match decomposition versus naive sliding-window (ablation) | Em-Garde · OVO-Bench Forward Active Responding | F1: 21.0 -> 31.0 | Improved |
| add RL stage after SFT (lambda=1) (ablation) | IGPP · OVO-Bench Forward Active Responding | F1: 23.4 -> 31.0 | Improved |
| add false-positive penalty lambda=1 versus 0 (ablation) | IGPP RL · OVO-Bench Forward Active Responding | F1: 27.7 -> 31.0 | Improved |
| raise lambda from 1 to 1.5 (ablation) | IGPP RL · OVO-Bench Forward Active Responding | F1: 31.0 -> 19.3 | Hurt |
| add false-positive penalty lambda=0 to RL on StreamingBench PO (ablation) | IGPP RL · StreamingBench Proactive Output | accuracy: SFT 34.8 -> RL lambda=0 30.8 | Hurt |
| match training and test threshold at 0.04 versus 0.03 (ablation) | Em-Garde · OVO-Bench Forward Active Responding | Avg F1: 26.59 -> 30.99 | Improved |
| raise streaming fps from 1 to 2 (ablation) | Em-Garde · StreamingBench Proactive Output | accuracy: 37.6 -> 38.0 | Improved |
| use Em-Garde instead of MMDuet-2 on ProactiveVideoQA WEB | Em-Garde · ProactiveVideoQA WEB | PAUC: 53.3 -> 44.3 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- StreamForest — StreamForest: Efficient Online Video Understanding with Persistent Event Memory (NeurIPS 2025) · better on 5 of 6 shared results
- VideoLLM-Online — VideoLLM-online: Online Video Large Language Model for Streaming Video (CVPR 2024) · better on 2 of 2 shared results
- Dispider — Dispider: Enabling Video LLMs with Active Real-Time Interaction via Disentangled Perception, Decision, and Reaction (CVPR 2025) · better on 1 of 1 shared result
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
- StreamOV: Streaming Omni-Video Understanding via Evidence-Guided Memory and Response Triggering (NeurIPS 2026 · similarity 0.89)
- StreamGaze: Gaze-Guided Temporal Reasoning and Proactive Understanding in Streaming Videos (NeurIPS 2026 · similarity 0.88)
- What Should a Streaming Video Model Remember? (NeurIPS 2026 · similarity 0.88)
- StreamBridge: Turning Your Offline Video Large Language Model into a Proactive Streaming Assistant (NeurIPS 2025 · similarity 0.86)
- Eyes Wide Open: Ego Proactive Video-LLM for Streaming Video (NeurIPS 2025 · similarity 0.86)
- LiveStar: Live Streaming Assistant for Real-World Online Video Understanding (NeurIPS 2025 · 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.
