Reversing the Flow: Generation-to-Understanding Synergy in Large Multimodal Models

2026-05-15

AI Summary

Current large-scale multimodal models exhibit a unidirectional dependency in which visual understanding and generation operate in isolation, lacking mechanisms for generation to actively inform understanding. This work proposes a "Generation-to-Understanding" (G2U) framework that systematically explores and implements such reciprocal support for the first time. By leveraging controllable visual generation—such as detail enhancement, contextual expansion, and structural visualization—as intermediate reasoning steps, the approach establishes a closed-loop generation–understanding pipeline within a unified model. This enables self-generated visual feedback to refine perceptual capabilities without requiring model retraining or external tools. Evaluated across twelve benchmarks, the method consistently enhances multimodal understanding performance, revealing that “imagination is the starting point of understanding” and demonstrating that the fidelity of generated content sets an upper bound on perceptual gains. The study also highlights that while current models can produce plausible edits, they remain limited in task alignment and self-reflection.

Abstract

The long-standing goal of multimodal AI is to build unified models in which visual understanding and visual generation mutually enhance one another. Despite recent works such as BAGEL, BLIP3o achieves remarkable progress; In practice, however, this unification remains one-directional: understanding routinely guides generation, yet how and why generation can support understanding is rarely investigated. We revisit this asymmetry and propose Generation-to-Understanding (G2U) synergy, where visual generation becomes an explicit intermediate reasoning step. Our framework enables a model to perform controlled generative acts, such as detail enhancement, context expansion or structural visualisation, to produce self-generated visual thoughts, which are then fed back into the model to refine perception without retraining or external tools. Through a comprehensive evaluation on twelve benchmarks, this reversed information flow consistently improves multimodal understanding. We show that generative fidelity bounds perceptual gain and that distinct families of edit prompts govern transfer efficiency. We further analyse whether models can decide what to imagine. While they can produce plausible edits, these self-generated visual thoughts lack stable task alignment, revealing that current large multimodal models fall short of true self-reflection. This work exposes a missing mechanism in unified cognition and suggests that imagination is not the end of understanding but its beginning.

Problem

• multimodal models
• visual understanding
• visual generation
• generation-to-understanding
• cognitive synergy

Innovation

• Generation-to-Understanding
• visual reasoning
• self-generated visual thoughts
• multimodal synergy
• controlled generation