VideoKR: Towards Knowledge- and Reasoning-Intensive Video Understanding

We introduce VideoKR, the first large-scale training corpus specifically designed to strengthen knowledge- and reasoning-intensive video understanding. It comprises 315K video reasoning examples over 145K newly collected, CC-licensed, expert-domain videos. We develop a human-in-the-loop, skill-oriented example generation pipeline that targets progressively deeper video reasoning capabilities while ensuring the difficulty, diversity, and reliability of both the examples and their CoT rationales. We also curate VideoKR-Eval, a new expert-annotated benchmark where questions require genuine video understanding and knowledge-intensive reasoning rather than textual shortcuts. Our experiments show that, under a standard SFT$\rightarrow$GRPO pipeline, models post-trained on VideoKR outperform prior post-training approaches on knowledge-intensive video reasoning while remaining competitive on general video reasoning, highlighting data design as a key driver of progress in video reasoning. We further conduct comprehensive ablations to isolate the contributions of VideoKR, providing actionable insights for future work.

Videos often require more than recognizing objects or reading subtitles: a person may need background knowledge and step-by-step thinking to understand what is happening. To help AI systems improve at this kind of video understanding, we build VideoKR, a large collection of video question-answer examples from expert-domain videos. These examples teach models to connect what they see in a video with relevant knowledge and reasoning. We also create a new evaluation set to test whether models truly understand videos rather than relying on simple text clues. Experiments show that training with VideoKR helps models answer more challenging video questions, especially those that require knowledge and reasoning.