UniEvo-VL: An On-policy Self-Distillation Training Recipe for Multimodal Model Self-improvement
UniEvo-VL improves multimodal image generation via self-distillation that minimizes divergence between student and critique-conditioned teacher diffusion distributions during self-correction. Experiments on Qwen2.5-Image improve GenEval scores from 0.747 to 0.808 without external teachers.
Published Sep 30, 2026▲ 292 on Hugging FacearXiv ↗
Only vote on papers you've read. Sign in with GitHub to vote.
UniEvo-VL delivers a clever on-policy self-distillation recipe that lifts GenEval scores via self-critique as privileged information, though uneven text-rendering gains and missing variance leave its universal self-improvement ceiling unproven.
Abstract
Modern multimodal models bring generation and understanding into a single unified system, which enables them to provide and learn from their own feedback. Motivated by this unified capacity, we introduce UniEvo-VL, a self-evolving framework for multimodal models to learn from this constructive self-correction feedback during test-time compute. Instead of relying on a separate, often larger, teacher, we leverage their self-critiques as privileged information and ask a single multimodal model to act as both teacher and student with different contexts. The student only sees the vanilla question, while the teacher conditions on the privileged critique. Then training minimizes the per-state divergence between their denoising diffusion distributions over the student's own sampling trajectories. Experiments demonstrate that UniEvo-VL improves the image generation capabilities of multimodal models, while maintaining their sensitivity to additional reflection information. Specifically, we build on top of the open-source Qwen-image-2512 and observe a significant performance gain from 0.747 to 0.808 on GenEval and from 32.97 to 35.53 on GenEval2 Soft-TIFA. Moreover, attempts with more powerful external critics (e.g., GPT5.6-Luna) show that multimodal models with strong judge capabilities can anticipate a higher self-evolving ceiling. Last but not least, mixed text-rendering outcomes show that our self-improvements may not be uniform across different tasks. Our study aims to shed light on the current hot recursive self-improvement research line to enhance the user experience when using multimodal models without external supervision or guidance.