arXiv cs.AI· Junming Liu, Yuqi Li, Yifei Sun, Maonan Wang, Yiming Cheng, Rui Qian, Piotr Koniusz, Yirong Chen, Ding Wang·· 9 天前AI 评分23
SAGE：通过几何逻辑一致性实现视觉语言模型自演化空间推理
Self-Evolving Spatial Reasoning in Vision Language Models via Geometric Logic Consistency
AI 导读
研究者提出 Spatial Alignment via Geometric Evolution（SAGE），一个通过几何与语言对偶操作提升视觉语言模型鲁棒空间推理的自演化框架。SAGE 将对偶一致性引入 GRPO 训练，并协同演化对偶生成与求解，配合动态操作池聚焦高信息量信号。该框架与模型无关、数据高效，可作为轻量适配阶段应用于任意现有 VLM，在视频与空间推理基准上持续优于强基线。
正文 · 原文
Abstract:Vision-Language Models (VLMs) have made striking progress, yet their spatial reasoning remains fragile. Models that answer an original input correctly can still fail under valid transformations with predictable answer mappings, revealing a gap between instance-level correctness and robust spatial reasoning. To address this, we propose Spatial Alignment via Geometric Evolution (SAGE), a self-evolving framework that improves robust spatial reasoning through geometric and linguistic duality operations. SAGE incorporates duality consistency into GRPO training, encouraging models to produce coherent answers across original and transformed inputs. SAGE co-evolves duality generation and solution, allowing the model to continually expose and address its own reasoning weaknesses. A dynamic operation pool identifies challenging operations and retires mastered ones, keeping training focused on informative duality signals. SAGE is model-agnostic, data-efficient compared to prior post-training methods, and can be applied as a lightweight adaptation stage to any existing VLM. Experiments on video and spatial reasoning benchmarks demonstrate consistent improvements over strong baselines and enhanced generalization to unseen data.
| Comments: | 36 pages, 7 figures, 14 tables |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2605.18162 [cs.CV] |
| (or arXiv:2605.18162v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2605.18162 arXiv-issued DOI via DataCite |
Submission history
From: Junming Liu [view email]
[v1]
Mon, 18 May 2026 10:05:21 UTC (14,921 KB)
[v2]
Thu, 8 Oct 2026 08:45:44 UTC (16,179 KB)
来源：arXiv cs.AI · arxiv.org