Abstract
Video multimodal large language models have made strong progress in open-ended video understanding, but still lack precise local spatiotemporal perception. When two videos share almost the same global semantics and differ only in a short time span or a small region, current models often fail to identify the change and provide reliable evidence.
We propose DELTAVID, a verifiable proxy-task framework that enhances fine-grained spatiotemporal perception with cross-video differences. The key idea is to turn cross-video spot-the-difference into a trainable perception signal, where a model identifies local changes, judges temporal boundaries, and organizes spatial evidence by comparing similar videos.
To make this signal scalable for training and reliable for evaluation, we further introduce DELTAVID-10K and DELTAVID-Bench, which convert controllable local differences in real videos into evidence-labeled training and test samples.
Experiments show that DELTAVID substantially improves performance in cross-video difference understanding and transfers the learned local evidence ability to general video understanding benchmarks, including MMVU, MLVU, Video-MME, VideoHolmes, VideoMMMU, LVBench, TempCompass, and LongVideoBench.
These results demonstrate that cross-video differences are not only an effective way to diagnose fine-grained perception failures but also a scalable proxy supervision that advances Video MLLMs from coarse semantic understanding towards fine-grained spatiotemporal evidence reasoning.
Blogger's Review: DELTAVID framework significantly enhances models' capabilities in spatiotemporal perception through innovative cross-video analysis. This advancement not only improves accuracy in video understanding but also opens new research avenues for future multimodal models.