Research map / Video / Video-language understanding
SVCBench: A Streaming Video Counting Benchmark for Spatial-Temporal State Maintenance
Pengyiang Liu, Zhongyue Shi, Hongye Hao, Qi Fu, Xueting BI, Siwei Zhang, Xiaoyang Hu, Zitian Wang, Linjiang Huang, Si 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 |
|---|---|---|---|
| overlay explicit cycle count annotation on video (ablation) | Gemini-3-Flash · SVCBench E2-Periodic (56 TOMATO questions) | GPA: 3.9 -> 81.8 | Improved |
| overlay explicit cycle count annotation on video (ablation) | Gemini-3-Flash · SVCBench E2-Periodic (56 TOMATO questions) | MoC: violations 43.8% -> 6.2% | Improved |
| use natural long-period videos instead of loop-concatenated (ablation) | Gemini-3-Flash · SVCBench E2-Periodic | GPA: 3.9 -> 2.9 | No clear effect |
| use natural long-period videos instead of loop-concatenated (ablation) | Doubao-Seed-1.8 · SVCBench E2-Periodic | GPA: 0.8 -> 0.5 | No clear effect |
| increase rotation cycles from first 10 to 60 (ablation) | Gemini-3-Flash · Rotating-camera bench identity persistence | GPA: 76.2 -> 28.3 | Hurt |
| increase rotation cycles from first 10 to 60 (ablation) | Doubao-Seed-1.8 · Rotating-camera bench identity persistence | GPA: 71.8 -> 35.7 | Hurt |
| increase rotation cycles from first 10 to 60 (ablation) | Qwen3-VL-8B · Rotating-camera bench identity persistence | GPA: 68.5 -> 23.8 | Hurt |
| widen GPA tolerance sigma from 0.05 to 0.20 (ablation) | Gemini-3-Flash · SVCBench overall | GPA: 37.0 -> 43.2 | No clear effect |
| raise frame sampling rate from 0.5 to 4 fps (ablation) | Qwen2.5-VL-7B · SVCBench E1-Action | GPA: 9.3 -> 10.2 | No clear effect |
| raise frame sampling rate from 0.5 to 4 fps (ablation) | Qwen2.5-VL-7B · SVCBench E2-Periodic | GPA: 0.0 -> 0.1 | No clear effect |
| use online streaming compression instead of offline backbone (second half of trajectory) (ablation) | StreamingVLM vs Qwen2.5-VL-7B · SVCBench overall | GPA: 24.5 -> 13.5 vs 21.5 -> 17.0 | Hurt |
| move from early to late third of video (ablation) | Gemini-3-Flash · SVCBench O2-Unique | GPA: 42.3 -> 29.1 | Hurt |
Closest papers by meaning
- Open-World Object Counting in Videos (AAAI 2026 · similarity 0.87)
- WorldReasonBench: Human-Aligned Stress Testing of Video Generators as Future World-State Predictors (NeurIPS 2026 · similarity 0.86)
- Video World Models with Long-term Spatial Memory (NeurIPS 2025 · similarity 0.86)
- What Should a Streaming Video Model Remember? (NeurIPS 2026 · similarity 0.85)
- Learning Object State Changes in Videos: An Open-World Perspective (CVPR 2024 · similarity 0.85)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · 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.
