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
  • 被引:0
  • 被引来源:Semantic Scholar + OpenAlex
  • S2被引:0
  • OpenAlex被引:0
  • 影响力被引:0
  • TLDR:FactoSR is presented, a factorized reinforcement learning framework that explicitly interpret the dimensions collapsed by visual projection, and suggests that reinforcing explicit, factorized 4D consistency is a critical step toward evolving VLMs into robust, world-aware reasoners.
  • OpenAlex ID:W7208725167
  • OpenAlex DOI:10.48550/arxiv.2609.03729
  • DOI:10.48550/arxiv.2609.03729
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://arxiv.org/pdf/2609.03729
  • OpenAlex更新:2026-09-10
  • 待LLM分类:否
  • 标题中文:展开世界:在强化空间推理中分解 4D 属性
  • TLDR中文:提出 FactoSR,一个因子化强化学习框架,显式解读视觉投影所塌缩的维度,并指出强化显式的、因子化的 4D 一致性是将 VLM 演化为稳健、具有世界感知能力的推理器的关键一步。
  • 来源文件:
  • /inbox/tom/_candidates/2026-09-08-rag-retrieval-reranking-candidates.json
  • /inbox/tom/_candidates/2026-09-08-agent-rag-longcontext-candidates.json
  • [OpenAlex backfill]
  • [S2 enrich]