arXiv：cs.AI· Wen Ye, Yuxiao Qu, Aviral Kumar, Xuezhe Ma·· 13 小时前
MIRROR：用跨视图互学提升多模态推理
MIRROR: Learning from the Other View for Multi-Modal Reasoning
AI 导读
针对视觉语言模型在文本与图像两种呈现下推理表现不一致的问题，研究者构建了配对多模态几何数据集 ODA-Data，并提出强化学习方法 MIRROR，从模型自身不同模态能力中提取跨视图自蒸馏信号：对每道题在所有视图下评估、选最优视图作教师，用带 reverse-KL 正则项的 GRPO 任务奖励目标训练其余视图。在多模态数学推理基准上，MIRROR 优于标准 RL，跨输入模态表现更准确一致。
正文
Abstract:Vision-language models (VLMs) can behave very differently depending on whether the same underlying problem is presented textually or visually. This inconsistency suggests that different views expose different reasoning paths and failure modes that standard multimodal post-training does not fully exploit. To study and exploit this phenomenon, we construct ODA-Data, a high-quality paired multimodal geometry dataset with text-dominant, image-dominant, and combined image+text views of the same problems, together with splits for training and evaluating view dependent reasoning behaviors. We then develop Modality-Informed Reciprocal Reasoning Optimization (MIRROR), a reinforcement learning approach that derives a cross-view self-distillation signal from the model's own modality-dependent capabilities. For each problem, MIRROR evaluates the model under all views, selects the best-performing view as a teacher, and trains the remaining views using a GRPO task-reward objective augmented with a reverse-KL regularizer toward that teacher. Across reasoning benchmarks that evaluate on multimodal mathematics problems, MIRROR improves over standard RL and yields more accurate and consistent behavior across different input modality representations.
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.21552 [cs.AI] |
| (or arXiv:2607.21552v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.21552 arXiv-issued DOI via DataCite |
Submission history
From: Wen Ye [view email]
[v1]
Thu, 23 Jul 2026 17:35:56 UTC (1,531 KB)
[v2]
Thu, 8 Oct 2026 18:38:17 UTC (1,514 KB)
来源：arXiv：cs.AI · arxiv.org