Paper page - Qwen-Image-2.0-RL Technical Report

Source: https://huggingface.co/papers/2606.27608 Published on Jun 25

#2 Paper of the day Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

A reinforcement learning and on-policy distillation approach enhances the visual quality and instruction-following capabilities of a diffusion model for image generation and editing tasks.

We present Qwen-Image-2.0-RL, a post-training pipeline that appliesreinforcement learning from human feedback(RLHF) andon-policy distillation(OPD) to improve both the visual quality and instruction-following capability of the Qwen-Image-2.0 diffusion model. To provide reliable reward signals, we construct task-specific composite reward models by fine-tuningvision-language modelswith apointwise scoring paradigmandchain-of-thought reasoning. Fortext-to-image generation, the reward models cover alignment, aesthetics, and portrait fidelity dimensions. Forimage editingtasks, the reward system addresses instruction-following accuracy and face identity preservation. Building on this reward system, we develop a scalableGRPO-based RL training framework, incorporating ahybrid classifier-free guidance(CFG) strategy to preserve pre-trained knowledge, prompt curation viaintra-group reward range filtering, andper-category reward weight calibration. To merge the task-specialized RL policies for T2I and editing, we proposeon-policy distillationas the final training stage, which consolidates multiple teachers into a single student model throughtrajectory-level velocity matching. Extensive evaluation shows that Qwen-Image-2.0-RL achieves 57.84 overall score on Qwen-Image-Bench (+2.61 over the base model), Elo ratings of 1193 in text-to-image arena (+78) and 1349 in image edit arena (+93), demonstrating consistent gains in aesthetic quality, prompt adherence, and editing accuracy.

View arXiv pageView PDFAdd to collection

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2606.27608 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2606.27608 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2606.27608 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.