MM-Zero: Self-Evolving Multi-Model Vision Language Models From Zero Data

2026-03-10

AI Summary

This work addresses the challenge of enabling vision-language models (VLMs) to undergo self-evolutionary training without any initial data. The authors propose MM-Zero, a novel framework that introduces a tripartite collaborative mechanism comprising a Proposer, a Coder, and a Solver. The Proposer generates abstract visual concepts and formulates questions, which the Coder translates into executable code (e.g., Python or SVG) to render corresponding images; the Solver then performs multimodal reasoning on these synthesized examples. Departing from conventional dual-agent paradigms, MM-Zero integrates Group Relative Policy Optimization (GRPO), visual verification, and a difficulty-balanced reward mechanism to achieve end-to-end self-evolution from scratch. Experiments demonstrate that MM-Zero substantially enhances VLM reasoning performance across multiple multimodal benchmarks, offering a scalable pathway toward autonomous model evolution.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Self-evolving has emerged as a key paradigm for improving foundational models such as Large Language Models (LLMs) and Vision Language Models (VLMs) with minimal human intervention. While recent approaches have demonstrated that LLM agents can self-evolve from scratch with little to no data, VLMs introduce an additional visual modality that typically requires at least some seed data, such as images, to bootstrap the self-evolution process. In this work, we present Multi-model Multimodal Zero (MM-Zero), the first RL-based framework to achieve zero-data self-evolution for VLM reasoning. Moving beyond prior dual-role (Proposer and Solver) setups, MM-Zero introduces a multi-role self-evolving training framework comprising three specialized roles: a Proposer that generates abstract visual concepts and formulates questions; a Coder that translates these concepts into executable code (e.g., Python, SVG) to render visual images; and a Solver that performs multimodal reasoning over the generated visual content. All three roles are initialized from the same base model and trained using Group Relative Policy Optimization (GRPO), with carefully designed reward mechanisms that integrate execution feedback, visual verification, and difficulty balancing. Our experiments show that MM-Zero improves VLM reasoning performance across a wide range of multimodal benchmarks. MM-Zero establishes a scalable path toward self-evolving multi-model systems for multimodal models, extending the frontier of self-improvement beyond the conventional two-model paradigm.

Problem

• zero-data
• self-evolving
• vision language models
• multimodal reasoning
• visual modality

Innovation

• zero-data self-evolution
• multi-role framework
• vision language models
• reinforcement learning
• executable code generation