Research map / Video / Video-language understanding
LongVT: Incentivizing "Thinking with Long Videos" via Native Tool Calling
Zuhao Yang, Sudong Wang, Kaichen Zhang, Keming Wu, Sicong Leng, Yifan Zhang, Bo Li, Chengwei Qin, Shijian Lu, Xingxuan Li, Lidong Bing
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 self-curated iMCoTT data to SFT (ablation) | LongVT-7B-SFT · avg of 4 benchmarks | average score: 24.8 -> 44.1 | Improved |
| add self-curated QAs to RL (ablation) | LongVT-7B-RL · VideoSIAH-Eval | score: 30.8 -> 35.9 | Improved |
| add SFT cold start before RL (ablation) | LongVT · avg of 4 benchmarks | average score: 41.9 -> 46.6 | Improved |
| add RL after SFT (ablation) | LongVT · avg of 4 benchmarks | average score: 44.1 -> 46.6 | Improved |
| add RFT after SFT+RL (ablation) | LongVT · VideoSIAH-Eval | score: 35.9 -> 42.0 | Improved |
| add RFT after SFT+RL (ablation) | LongVT · VideoMMMU adaptation / LVBench | score: adaptation 37.7 -> 35.7; LVBench 41.4 -> 41.3 | Mixed |
| use IoU reward instead of Recall for temporal grounding (ablation) | LongVT RL · Charades-STA | mIoU: 21.6 -> 27.2 | Improved |
| use Recall reward instead of no decoupled reward (ablation) | LongVT RL · Charades-STA | mIoU: 21.2 -> 21.6 | No clear effect |
| add tool-call reward (ablation) | LongVT RL · tool usage / accuracy | accuracy | No clear effect |
| apply tool calling and RL to LongVT vs Qwen2.5-VL-7B at 64 frames | LongVT-7B-RL · avg of 4 benchmarks (sparse) | average score: 37.2 -> 43.7 | Improved |
Closest papers by meaning
- A Benchmark for Omni-Modal Reasoning in Long Videos (NeurIPS 2026 · similarity 0.91)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.91)
- When Thinking Drifts: Evidential Grounding for Robust Video Reasoning (NeurIPS 2025 · similarity 0.90)
- SciVideoBench: Benchmarking Scientific Video Reasoning in Large Multimodal Models (ICML 2026 · similarity 0.90)
- OneThinker: All-in-one Reasoning Model for Image and Video (CVPR 2026 · similarity 0.90)
- MMR-V: What's Left Unsaid? A Benchmark for Multimodal Deep Reasoning in Videos (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.
