VLM-as-perceiver 范式将遮挡场景下的规划成功率最高提升至 97%
Isolating Perception for Multi-View Planning under Occlusion-Induced Partial Observability
视觉规划基准快被 VLM 刷满时,这篇论文找出了它们在遮挡场景下的真实短板:三个模型在 BW-PO 上接近全挂。把 VLM 只当感知器、推理交给外部求解器后成功率能到 84%-97%,做多视角规划或智能体开发的都该看看这套拆分思路。
一篇 arXiv 论文针对视觉语言模型在遮挡导致的局部可观测(PO)场景下深度歧义严重、随问题规模增大性能急剧下降的问题,发布了基于 Blocksworld 领域的受控基准 BW-PO。该基准让现有 VLM 规划器和形式化方法的表现降至接近零。论文提出 VLM-as-perceiver 范式:把 VLM 的职责收窄到局部感知,跨视角证据调和与计划生成则交给外部求解器完成。
在 BW-PO 各模型的难例子集上,这一改动让 Gemma-4-31B 的规划成功率从 1% 提升到 97%,Qwen3.5-122B-A10B 从 0% 提升到 95%,Gemini-3.1-Pro 从 3% 提升到 84%。论文认为这样把原本纠缠的两个瓶颈分开了:深度歧义下的证据调和,以及高密度场景下视觉抽取的完整性。
Isolating Perception for Multi-View Planning under Occlusion-Induced Partial Observability
Vision-language models (VLMs) are beginning to saturate visual planning benchmarks, yet they fail to resolve depth ambiguity and degrade sharply with problem size under occlusion-induced partial observability (PO). We introduce BW-PO based on the Blocksworld domain, a controlled benchmark that reduces existing VLM planners and formalizers to near-zero performance, and propose VLM-as-perceiver, a paradigm that narrows the VLM's role to local perception while delegating cross-view evidence reconciliation and plan generation to external solvers. Across model-specific challenging subsets of BW-PO, this raises plan success from 1% to 97% for Gemma-4-31B, 0% to 95% for Qwen3.5-122B-A10B, and 3% to 84% for Gemini-3.1-Pro. Isolating perception in this way separates two bottlenecks that are otherwise entangled: reconciliation under depth ambiguity, and completeness of visual extraction under scene density.
本条情报归属 AITOP「论文」核心追踪栏目。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:VLM、BW-PO、Gemma、Qwen,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明:本文原始线索采编自「arXiv cs.AI」(@Kevin Nguyen, Ceyhun Efe Kayan, Feng Liu, Li Zhang),经 AITOP 多维评分管线结构化提炼。