GenDSR：从视频生成模型迁移表征以提升动态空间推理
GenDSR: Transferring Representations from Video Generation Models for Dynamic Spatial Reasoning
AI 导读
GenDSR 框架将视频生成模型（VGM）的中间表征迁移到视觉语言模型（VLM），用于动态空间推理。该框架设计原始分支与差分分支分别保留整体生成上下文与时间变化信息，并通过任务条件化的 token 级门控融合注入原有视觉表征。在 DSR-Bench 上 GenDSR 达到 67.2%，超越此前最佳成绩，创下新的 SOTA。
正文 · 原文
Abstract:Dynamic spatial reasoning is essential for connecting visual intelligence to thephysical world, yet remains challenging for vision-language models (VLMs), which must preserve scene state while tracking how spatial relations evolve over time. Meanwhile, video generation models (VGMs), trained to synthesize temporally coherent world dynamics, offer a rich source of spatiotemporal representations. We present GenDSR, a framework that transfers intermediate representations from VGM into VLM for dynamic spatial reasoning. Rather than treating a VGM feature as a monolithic prior, GenDSR constructs two complementary branches: a raw branch preserving holistic generative context and a delta branch emphasizing temporal variation. Task-conditioned tokenwise gates then fuse both signals and inject the result into the native visual representation. On DSR-Bench, GenDSR achieves 67.2%, surpassing the previous best result and establishing a new state of the art.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.41314 [cs.CV] |
| (or arXiv:2610.41314v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.41314 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Ke Yang [view email]
[v1]
Fri, 30 Oct 2026 06:26:31 UTC (152 KB)
来源：arXiv cs.AI · arxiv.org