Research map / Video / Video-language understanding
Conan: Progressive Learning to Reason Like a Detective over Multi-Scale Visual Evidence
Kun Ouyang, Yuanxin Liu, Linli Yao, Yishuo Cai, Hao Zhou, Fandong Meng, Jie Zhou, Xu Sun
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 multi-scale (context) frame categorization (ablation) | Conan · six video reasoning benchmarks (avg) | accuracy: 53.8 -> 57.4 | Improved |
| Add evidence difficulty-aware sampling (ablation) | Conan · six video reasoning benchmarks (avg) | accuracy: 55.2 -> 57.4 | Improved |
| Add textual reasoning cold-start stage (ablation) | Conan · six video reasoning benchmarks (avg) | accuracy: 57.0 -> 57.4 | Improved |
| Add multimodal alignment stage (ablation) | Conan · six video reasoning benchmarks (avg) | accuracy: 56.4 -> 57.4 | Improved |
| Add vision-centric reasoning stage (ablation) | Conan · six video reasoning benchmarks (avg) | accuracy: 53.0 -> 57.4 | Improved |
| Add multi-stage cold-start before RLVR (ablation) | Conan · six video reasoning benchmarks (avg) | accuracy: 51.0 -> 57.4 | Improved |
| Add identification reward (ablation) | AIR RLVR · six video reasoning benchmarks (avg) | accuracy: 53.8 -> 57.4 | Improved |
| Add retrieval reward (ablation) | AIR RLVR · six video reasoning benchmarks (avg) | accuracy: 54.0 -> 57.4 | Improved |
| Use Conan-style evidence retrieval instead of single-round text CoT (ablation) | Conan · six video reasoning benchmarks (avg) | accuracy: 55.2 -> 57.4 | Improved |
| Skip textual reasoning cold-start stage (ablation) | Conan · VRBench | accuracy: 81.4 vs 81.0 | No clear effect |
| Train Conan from Qwen2.5-VL-7B-Instruct | Qwen2.5-VL-7B-Instruct · Video-Holmes | accuracy: 28.5 -> 44.6 | Improved |
| Train Conan from Qwen2.5-VL-7B-Instruct | Qwen2.5-VL-7B-Instruct · long video understanding (avg of four) | accuracy: 48.0 -> 54.9 | Improved |
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.
- Video-MTR — Video-MTR: Reinforced Multi-Turn Reasoning for Long Video Understanding (ICML 2026) · better on 6 of 6 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
- When Thinking Drifts: Evidential Grounding for Robust Video Reasoning (NeurIPS 2025 · similarity 0.90)
- MVVBench: Benchmarking 4D Reasoning in Vision-Language Models (NeurIPS 2026 · similarity 0.88)
- Look, Remember and Reason: Grounded Reasoning in Videos with Language Models (ICLR 2024 · similarity 0.88)
- SLVR: Structured Latent Visual Reasoning via Human-like Reasoning Flows (NeurIPS 2026 · similarity 0.88)
- DeepScan: A Training-Free Framework for Visually Grounded Reasoning in Large Vision-Language Models (CVPR 2026 · similarity 0.88)
- See It, Say It, Sorted: An Iterative Training-Free Framework for Visually-Grounded Multimodal Reasoning in LVLMs (CVPR 2026 · similarity 0.88)
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.
