Mimic Human Cognition, Master Multi-Image Reasoning: A Meta-Action Framework for Enhanced Visual Understanding

TL;DR

This paper introduces CINEMA, a human cognition-inspired framework for multi-image reasoning that decomposes reasoning into structured meta-actions, improving performance across various visual understanding benchmarks.

Contribution

The paper presents a novel meta-action framework inspired by human cognition, along with a retrieval-based training strategy and a two-stage reinforcement learning paradigm for enhanced multi-image reasoning.

Findings

Achieved state-of-the-art results on multiple multi-image reasoning benchmarks.

Surpassed GPT-4o on MUIR and MVMath benchmarks.

Outperformed specialized video reasoning models on video understanding tasks.

Abstract

While Multimodal Large Language Models (MLLMs) excel at single-image understanding, they exhibit significantly degraded performance in multi-image reasoning scenarios. Multi-image reasoning presents fundamental challenges including complex inter-relationships between images and scattered critical information across image sets. Inspired by human cognitive processes, we propose the Cognition-Inspired Meta-Action Framework (CINEMA), a novel approach that decomposes multi-image reasoning into five structured meta-actions: Global, Focus, Hint, Think, and Answer which explicitly modeling the sequential cognitive steps humans naturally employ. For cold-start training, we introduce a Retrieval-Based Tree Sampling strategy that generates high-quality meta-action trajectories to bootstrap the model with reasoning patterns. During reinforcement learning, we adopt a two-stage paradigm: an exploration phase with Diversity-Preserving Strategy to avoid entropy collapse, followed by an annealed exploitation phase with DAPO to gradually strengthen exploitation. To train our model, we construct a dataset of 57k cold-start and 58k reinforcement learning instances spanning multi-image, multi-frame, and single-image tasks. We conduct extensive evaluations on multi-image reasoning benchmarks, video understanding benchmarks, and single-image benchmarks, achieving competitive state-of-the-art performance on several key benchmarks. Our model surpasses GPT-4o on the MUIR and MVMath benchmarks and notably outperforms specialized video reasoning models on video understanding benchmarks, demonstrating the effectiveness and generalizability of our human cognition-inspired reasoning framework.