Research map / Video / Video-language understanding
Understanding Temporal Logic Consistency in Video-Language Models through Cross-Modal Attention Discriminability
Chengzhi Li, Heyan Huang, Ping Jian, Zhen Yang, Yaning Tian, Zhongbin Guo
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 TCAS loss to SFT | TimeChat (VTune) · Charades-CON | Ground: 76.2 -> 83.3 | Improved |
| add TCAS loss to SFT | TimeChat (VTune) · Charades-CON | H-Verify: 44.8 -> 52.9 | Improved |
| add TCAS loss to SFT | TimeChat (TimeIT) · Charades-CON | S-Ground: 10.5 -> 33.8 | Improved |
| add TCAS loss to SFT | Qwen2.5-VL-7B (VTune) · Charades-CON | Ground: 28.3 -> 34.0 | Improved |
| add TCAS loss to SFT | TimeChat (VTune) · ActivityNet-CON | Ground: 37.4 -> 38.8 | Improved |
| add TCAS loss to SFT | TimeChat · Charades-STA | R@1,0.7: 34.7 -> 37.6 | Improved |
| add TCAS loss to SFT | Video-LLaMA-7B · Charades-STA | R@1,0.7: 20.1 -> 20.1 | No clear effect |
| add TCAS loss to SFT | Qwen2.5-VL-7B · ActivityNet Captions | R@1,0.5: 16.5 -> 16.5 | No clear effect |
| add TCAS loss to SFT on videos longer than 40s | TimeChat (VTune) · Charades-CON, videos >40s | Ground: 58.8 -> 76.5 | Improved |
| intervene mildly on attention toward ground truth (alpha 0.2) at inference (ablation) | TimeChat · Charades-CON rephrased grounding | R0.7: 79.9 -> 80.8 | Improved |
| intervene strongly on attention (alpha 1.0) at inference (ablation) | TimeChat · Charades-CON grounding | mIoU: 65.8 -> 63.2 | Hurt |
| random attention intervention alpha 0.1 (ablation) | TimeChat SFT · Charades-CON | c_rg: 65.4 -> 2.1 | Hurt |
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 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
- On the Consistency of Video Large Language Models in Temporal Comprehension (CVPR 2025 · similarity 0.92)
- SlowFocus: Enhancing Fine-grained Temporal Understanding in Video LLM (NeurIPS 2024 · similarity 0.91)
- Enhancing Temporal Understanding in Video-LLMs through Stacked Temporal Attention in Vision Encoders (NeurIPS 2025 · similarity 0.90)
- MMBench-Video: A Long-Form Multi-Shot Benchmark for Holistic Video Understanding (NeurIPS 2024 · similarity 0.89)
- Temporal Reasoning Transfer from Text to Video (ICLR 2025 · similarity 0.89)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.89)
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.
