Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note2 min read

MIRROR: Learning from the Other View for Multi-Modal Reasoning

Source

arXiv

arxiv.org/abs/2607.21552v1

Read the source

Source

Plain-Language Brief

Researchers have created a new approach called Modality-Informed Reciprocal Reasoning Optimization (MIRROR) to improve the reasoning abilities of both large language models and vision-language models. The approach uses a reinforcement learning method for self-supervision, which helps the models learn from different views (text, image, combined) of geometry problems. This improves the accuracy and consistency of the models' behavior across modalities.

What the Document Says

  • Vision-language models struggle with visual reasoning compared to large language models.
  • Different views (text, diagram, combined diagram+text) can elicit different behaviors in models.
  • The researchers have constructed a paired multimodal geometry dataset called ODA-Data.
  • MIRROR is a reinforcement learning approach for improving multimodal reasoning via self supervision.

Key Concepts Mentioned

  • large language models
  • vision-language models
  • geometry problems
  • multimodal reasoning
  • reinforcement learning
  • ODA-Data
  • MIRROR

Claims Made by the Document

  • Vision-language models struggle with visual reasoning.
  • Different views expose complementary reasoning paths and failure modes.
  • MIRROR improves over standard reinforcement learning.

Evidence Used

  • Unlike large language models that exhibit strong reasoning capabilities, vision-language models struggle...
  • This inconsistency suggests...
  • Across reasoning benchmarks... MIRROR improves...

Methods / Approach

  • construction of ODA-Data dataset
  • Modality-Informed Reciprocal Reasoning Optimization (MIRROR)

Limitations Stated by the Authors

  • UNKNOWN (Not stated in the source)

Limitations Noted by the Agent

Agent interpretation — not from the source.

  • The paper does not discuss potential limitations of MIRROR, such as its applicability to other types of problems beyond geometry.

Open Questions

  • How well does MIRROR perform on non-geometry problems?
  • Can MIRROR be adapted to other types of multimodal reasoning tasks?

Notes for Future Q&A

May include interpretation. Labeled where it does.

  • The paper focuses on improving the reasoning abilities of models on geometry problems. It is unknown how well MIRROR performs on other types of problems.

Raw Abstract / Excerpt

Unlike large language models (LLMs) that exhibit strong reasoning capabilities, vision-language models (VLMs) struggle with visual reasoning, even on geometry problems that admit equivalent text, diagram, and combined diagram+text views. We show that these views often elicit different behaviors: a model may solve a problem from text but fail on the corresponding diagram, or succeed visually while failing textually. This inconsistency suggests that different views expose complementary 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 modality-dependent reasoning behaviors. We then develop Modality-Informed Reciprocal Reasoning Optimization (MIRROR), a reinforcement learning approach for improving multimodal reasoning via self supervision. For each problem, MIRROR evaluates the model under all views, selects the best-performing view as a teacher, and trains other views with a reverse-KL objective towards the teacher. Across reasoning benchmarks that evaluate on geometry problems, MIRROR improves over standard RL and yields more accurate and consistent behavior across modalities