Research map / Video / Video-language understanding
VideoSetDiff: Identifying and Reasoning Similarities and Differences in Similar Videos
Yue Qiu, Yanjun Sun, Takuma Yagi, Shusaku Egami, Natsuki Miyata, Ken Fukuda, Kensho Hara, Ryusuke Sagawa
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 |
|---|---|---|---|
| fine-tune with multi-video training set and cross-video modelling (VidSetReasoner vs InternVideo2.5) | InternVideo2.5 · VideoSetDiff evaluation set | overall accuracy: 66.8 -> 68.6 | Improved |
| fine-tune with multi-video training set (reasoning questions) | InternVideo2.5 · VideoSetDiff evaluation set | reasoning question accuracy: 64.8 -> 66.5 | Improved |
| fine-tune with multi-video training set (recognition questions) | InternVideo2.5 · VideoSetDiff evaluation set | recognition question accuracy: 68.7 -> 70.6 | Improved |
| add captions + GPT-4 instead of blind GPT-4 | Blind GPT-4 · VideoSetDiff evaluation set | overall accuracy: 51.7 -> 56.7 | Improved |
| use direct multi-video QA instead of caption + GPT-4 | VideoChat2 · VideoSetDiff evaluation set | overall accuracy: 56.7 -> 52.4 | No clear effect |
| use direct multi-video QA instead of caption + GPT-4 | InternVideo2.5 · VideoSetDiff evaluation set | overall accuracy: 56.9 -> 66.8 | Improved |
| use direct multi-video QA instead of caption + GPT-4 | TimeChat · VideoSetDiff evaluation set | overall accuracy: 57.1 -> 32.7 | Hurt |
| use multiple-choice instead of binary questions | VidSetReasoner · VideoSetDiff evaluation set | accuracy: 74.2 -> 62.7 | Hurt |
| use caption + GPT-4 pipeline instead of direct for recognition | InternVideo2 · VideoSetDiff evaluation set | recognition question accuracy: 59.3 -> 53.7 | Hurt |
| use caption + GPT-4 instead of direct QA for reasoning | InternVideo2 · VideoSetDiff evaluation set | reasoning question accuracy: 45.6 -> 60.9 | Improved |
| increase number of input frames (ablation) | VideoLLaMA2 and GPT4o · VideoSetDiff evaluation set | accuracy | No clear effect |
| increase number of input frames (ablation) | InternVideo2.5 · VideoSetDiff evaluation set | accuracy | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TimeChat — TimeChat: A Time-sensitive Multimodal Large Language Model for Long Video Understanding (CVPR 2024) · better on 3 of 3 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
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.88)
- Video Action Differencing (ICLR 2025 · similarity 0.87)
- ConViS-Bench: Estimating Video Similarity Through Semantic Concepts (NeurIPS 2025 · similarity 0.87)
- SF2T: Self-supervised Fragment Finetuning of Video-LLMs for Fine-Grained Understanding (CVPR 2025 · similarity 0.86)
- Improving LLM Video Understanding with 16 Frames Per Second (ICML 2025 · similarity 0.86)
- Storyboard-guided Alignment for Fine-grained Video Action Recognition (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.
