CamReasoner: Reinforcing Camera Movement Understanding via Structured Spatial Reasoning

Understanding camera dynamics is a fundamental pillar of video spatial intelligence. However, existing multimodal models predominantly treat this task as a black-box classification, often confusing physically distinct motions by relying on superficial visual patterns rather than geometric cues. We present $\mathbf{\text{CamReasoner}}$, a framework that reformulates camera movement understanding as a structured inference process to bridge the gap between perception and cinematic logic. Our approach centers on the Observation-Thinking-Answer (O-T-A) paradigm, which compels the model to articulate spatio-temporal observations and reason about motion patterns within an explicit reasoning block. To instill this capability, we construct a Large-scale Inference Trajectory Suite comprising 18k SFT reasoning chains and 38k RL feedback samples. To the best of our knowledge, $\mathbf{\text{we}}\mathbf{\text{are}}\mathbf{\text{the}}\mathbf{\text{first}}\mathbf{\text{to}}\mathbf{\text{employ}}\mathbf{\text{RL}}\mathbf{\text{for}}\mathbf{\text{logical}}\mathbf{\text{alignment}}\mathbf{\text{in}}\mathbf{\text{camera}}\mathbf{\text{movement}}\mathbf{\text{understanding}}$, ensuring motion inferences are grounded in structured visual reasoning rather than contextual guesswork. Built upon Qwen2.5-VL-7B, CamReasoner-7B improves binary classification accuracy from 73.8% to 78.4% and VQA accuracy from 60.9% to 74.5% over its backbone, consistently outperforming both proprietary and open-source baselines across multiple benchmarks.