Research map / Video / Video-language understanding
CrossVid: A Comprehensive Benchmark for Evaluating Cross-Video Reasoning in Multimodal Large Language Models
Jingyao Li, Jingyun Wang, Molin Tan, Haochen Wang, Cilin Yan, Likun Shi, Jiayin Cai, Xiaolong Jiang, Yao Hu
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 |
|---|---|---|---|
| increase input frames from 32 to 256 (ablation) | Qwen2.5-VL-72B · CrossVid overall | O.Avg accuracy: 33.8 -> 39.5 | Improved |
| increase input frames from 32 to 256 (ablation) | Qwen2.5-VL-72B · CrossVid free-form QA (CCQA) | accuracy: 18.9 -> 34.0 | Improved |
| increase input frames from 32 to 256 (ablation) | Qwen2.5-VL-72B · CrossVid temporal understanding | T.Avg accuracy: 33.8 -> 33.9 | No clear effect |
| add three-stage CoT prompt (ablation) | Qwen2.5-VL-72B · CrossVid overall | O.Avg accuracy: 34.4 -> 39.5 | Improved |
| add three-stage CoT prompt (ablation) | GPT-4.1 · CrossVid overall | O.Avg accuracy: 45.2 -> 44.9 | No clear effect |
| add three-stage CoT prompt (ablation) | MiniCPM-o 2.6 · CrossVid overall | O.Avg accuracy: 25.6 -> 23.7 | Hurt |
| add three-stage CoT prompt (ablation) | InternVL3-38B · CrossVid overall | O.Avg accuracy: 23.5 -> 24.4 | Mixed |
| thinking-enabled model vs non-thinking models | GLM-4.1V-9B-Thinking and MiMo-7B · CrossVid overall (<10B open-source) | O.Avg accuracy: 35.1 and 28.3 vs third best | Improved |
Closest papers by meaning
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.92)
- OmniVideoBench: Towards Audio-Visual Understanding Evaluation for Omni MLLMs (ICLR 2026 · similarity 0.92)
- MMR-V: What's Left Unsaid? A Benchmark for Multimodal Deep Reasoning in Videos (ICLR 2026 · similarity 0.92)
- MMOU: A Massive Multi-Task Omni Understanding and Reasoning Benchmark for Long and Complex Real-World Videos (NeurIPS 2026 · similarity 0.91)
- MMBench-Video: A Long-Form Multi-Shot Benchmark for Holistic Video Understanding (NeurIPS 2024 · similarity 0.91)
- VideoReasonBench: Can MLLMs Perform Vision-Centric Complex Video Reasoning? (ICLR 2026 · similarity 0.90)
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.
