Analyze-Prompt-Reason: A Collaborative Agent-Based Framework for Multi-Image Vision-Language Reasoning

TL;DR

This paper introduces a modular, automated framework using dual agents—a prompt generator and a vision-language model—for multi-image reasoning across diverse visual tasks, achieving high accuracy on multiple datasets.

Contribution

It proposes a training-free, generalizable agent-based framework for multi-image reasoning that outperforms existing methods on various visual question answering tasks.

Findings

LVLMs can effectively reason over multiple images with proper prompts

Claude 3.7 achieves near-ceiling performance on several challenging tasks

Design choices significantly impact reasoning performance

Abstract

We present a Collaborative Agent-Based Framework for Multi-Image Reasoning. Our approach tackles the challenge of interleaved multimodal reasoning across diverse datasets and task formats by employing a dual-agent system: a language-based PromptEngineer, which generates context-aware, task-specific prompts, and a VisionReasoner, a large vision-language model (LVLM) responsible for final inference. The framework is fully automated, modular, and training-free, enabling generalization across classification, question answering, and free-form generation tasks involving one or multiple input images. We evaluate our method on 18 diverse datasets from the 2025 MIRAGE Challenge (Track A), covering a broad spectrum of visual reasoning tasks including document QA, visual comparison, dialogue-based understanding, and scene-level inference. Our results demonstrate that LVLMs can effectively reason over multiple images when guided by informative prompts. Notably, Claude 3.7 achieves near-ceiling performance on challenging tasks such as TQA (99.13% accuracy), DocVQA (96.87%), and MMCoQA (75.28 ROUGE-L). We also explore how design choices-such as model selection, shot count, and input length-influence the reasoning performance of different LVLMs.