Summary
This paper introduces a self-evolving training framework that enables unified large multimodal models (LMMs) to improve both visual understanding and image generation autonomously, without curated post-training supervision such as human annotations, preference labels, or external reward models. The framework relies only on unlabeled images and assigns three internal roles to the model: a Proposer that generates visual questions, a Solver that answers and evaluates them, and a Generator that produces images. Improvement is driven by self-consistency rewards, allowing the model to refine both abilities in a closed loop. The work, authored by Ritesh Thawkar, Shravan Venkatraman, and Omkar Thawakar, is available on arXiv as 2606.27376.
Paper Overview
Research area: Computer Vision (CV)
Authors: Ritesh Thawkar, Shravan Venkatraman, Omkar Thawakar
Published: 2026-06-27
arXiv: 2606.27376
Abstract
Most unified large multimodal models (LMMs) that support both visual understanding and image generation still rely on curated post-training supervision, such as human annotations, preference labels, or external reward models. This paper asks whether a unified LMM can improve both abilities autonomously using only unlabeled images.
The authors propose a self-evolving training framework built on three internal roles:
- Proposer — generates visual questions from unlabeled images
- Solver — answers and evaluates the generated questions
- Generator — produces images, refined through the same feedback loop
Training is driven by
self-consistency rewards, so the model supervises its own improvement across both understanding and generation without external annotators or reward models.
---
*Automatically collected on 2026-06-27.*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178208189