Summary
AlphaGRPO is a framework that applies Group Relative Policy Optimization (GRPO) to AR-Diffusion Unified Multimodal Models (UMMs), enhancing multimodal generation without an additional cold-start stage. The approach unlocks two advanced capabilities: Reasoning Text-to-Image Generation, where the model infers implicit user intents, and Self-Reflective Refinement, where it autonomously diagnoses and corrects misalignments in its outputs. To provide stable supervision for real-world generation, the authors introduce Decompositional Verifiable Reward (DVReward), which uses an LLM to decompose complex user requests into atomic, verifiable semantic and quality questions that a general-purpose MLLM evaluates, yielding reliable and interpretable feedback. Experiments show robust gains on multimodal generation benchmarks including GenEval, TIIF-Bench, DPG-Bench, and WISE, plus significant improvements on GEdit editing tasks without any editing-specific training. Paper: arXiv 2605.12495 by Runhui Huang, Jie Wu, Rui Yang, Zhe Liu, and Hengshuang Zhao.
Overview
Field: Computer Vision
Authors: Runhui Huang, Jie Wu, Rui Yang, Zhe Liu, Hengshuang Zhao
Published: 2026-05-12
arXiv: 2605.12495
Summary
This paper proposes AlphaGRPO, a novel framework that applies Group Relative Policy Optimization (GRPO) to AR-Diffusion Unified Multimodal Models (UMMs) to enhance multimodal generation capabilities without an additional cold-start stage.
The approach unlocks the model's intrinsic potential to perform advanced reasoning tasks:
- Reasoning Text-to-Image Generation: the model actively infers implicit user intents.
- Self-Reflective Refinement: the model autonomously diagnoses and corrects misalignments in generated outputs.
To address the challenge of providing stable supervision for real-world multimodal generation, the authors introduce the
Decompositional Verifiable Reward (DVReward). Unlike holistic scalar rewards, DVReward utilizes an LLM to decompose complex user requests into atomic, verifiable semantic and quality questions, which are evaluated by a general-purpose MLLM to provide reliable and interpretable feedback.
Results
Extensive experiments demonstrate that AlphaGRPO achieves robust improvements on multimodal generation benchmarks including GenEval, TIIF-Bench, DPG-Bench, and WISE, and yields significant gains on GEdit editing tasks even without any editing-specific training.
---
*Auto-collected on 2026-05-14.*
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/177620001