EVE: Verifiable Self-Evolution of MLLMs via Executable Visual Transformations

TL;DR

EVE introduces a novel framework for self-evolving multimodal large language models by using executable visual transformations to generate verifiable training data, avoiding pseudo-labels and enabling continuous, diverse, and challenging model improvement.

Contribution

EVE presents a dual-policy architecture that synthesizes dynamic visual transformations with verified ground-truth answers, advancing self-evolution methods for MLLMs.

Findings

EVE outperforms existing self-evolution approaches in experiments.

The framework effectively maintains diversity and difficulty in training tasks.

EVE's approach ensures verifiable supervision without relying on model predictions.

Abstract

Self-evolution of multimodal large language models (MLLMs) remains a critical challenge: pseudo-label-based methods suffer from progressive quality degradation as model predictions drift, while template-based methods are confined to a static set of transformations that cannot adapt in difficulty or diversity. We contend that robust, continuous self-improvement requires not only deterministic external feedback independent of the model's internal certainty, but also a mechanism to perpetually diversify the training distribution. To this end, we introduce EVE (Executable Visual transformation-based self-Evolution), a novel framework that entirely bypasses pseudo-labels by harnessing executable visual transformations continuously enriched in both variety and complexity. EVE adopts a Challenger-Solver dual-policy architecture. The Challenger maintains and progressively expands a queue of visual transformation code examples, from which it synthesizes novel Python scripts to perform dynamic visual transformations. Executing these scripts yields VQA problems with absolute, execution-verified ground-truth answers, eliminating any reliance on model-generated supervision. A multi-dimensional reward system integrating semantic diversity and dynamic difficulty calibration steers the Challenger to enrich its code example queue while posing progressively more challenging tasks, preventing mode collapse and fostering reciprocal co-evolution between the two policies. Extensive experiments demonstrate that EVE consistently surpasses existing self-evolution methods, establishing a robust and scalable paradigm for verifiable MLLM self-evolution. The code is available at https://github.com/0001Henry/EVE .