Unfold The World: Factorize 4D Properties in Reinforcing Spatial Reasoning

  • 类型:arxiv
  • 标识:2609.03729
  • 链接:https://arxiv.org/abs/2609.03729
  • 主分类:multimodal
  • 形态:method
  • TLDR:Despite the remarkable prowess of Vision-Language Models (VLMs) in general multimodal tasks, they remain fundamentally flat'' when reasoning about the physical world. We argue that this spatial bottleneck stems from a profound dimensional mismatch: while VLMs are trained to interpret 2D projections, true spatial reasoning demands the recovery of latent 3D geometry and temporal continuity. To conquer this high-dimensional complexity, we advocate a shift from monolithic learning to adivide and conquer'' paradigm. We present FactoSR, a factorized reinforcement learning framework that explici
  • 待LLM分类:否
  • 来源文件
  • /inbox/tom/_candidates/2026-09-08-rag-retrieval-reranking-candidates.json