让视频模型自动选对推理方式，又准又省
Video-FLAIR: Not Whether to Reason, But How
- 用强化学习动态选择感知、组合或反思式推理
- 在多个数据集上准确率提升4.8~5.4点，耗 token 降至95
- 无需人工标注每题推理类型，适合复杂多模态任务
多模态问题需要不同类型的推理：有些可直接从视觉信号中提取信息（感知推理），有些需整合观察结果（组合推理），有些则需评估多个假设（反思推理）。但现有方法对所有问题采用统一推理策略，导致简单任务计算浪费，复杂任务推理不足。本文提出 Video-FLAIR 训练框架，利用强化学习让模型为每个问题自动选择合适的推理模式。训练时，模型对同一提示生成三种模式的响应，通过综合奖励机制比较其正确性、语义一致性与计算成本，选出最优解并提供监督信号。该方法无需逐题标注推理类型，即可实现自适应推理。在 MathVista、Video-Holmes 与 Video-MMMU 上，相比 Qwen2.5-VL 基线，准确率分别提升 +5.4、+4.8 和 +4.8，平均词元使用量从 417 降至 95。
原文摘要 · Abstract (English)
Multimodal queries can require different types of reasoning. Some can be answered via perceptual reasoning, extracting information directly from the visual signal, while others require compositional reasoning that combines observations or deliberative reasoning that evaluates competing hypotheses. However, many existing methods apply a uniform reasoning strategy across queries, leading to unnecessary computation on simple tasks and insufficient reasoning on complex ones. We introduce Video-FLAIR, a training framework that learns to select the appropriate reasoning mode for each query using reinforcement learning. During training, the model generates responses under all three modes for the same prompt, enabling direct comparison. A composite reward compares these responses to favor the most effective one based on correctness, grounding, and cost, while discouraging unsupported or misaligned deliberation. This yields a supervision signal for learning adaptive reasoning without per-query annotations. Video-FLAIR improves accuracy over the Qwen2.5-VL base model by +5.4 on MathVista, +4.8 on Video-Holmes, and +4.8 on Video-MMMU, while reducing average token usage to 95 compared to 417 for always-thinking baselines.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。