Research map / Image generation / Text-to-image diffusion
AlphaGRPO: Unlocking Self-Reflective Multimodal Generation in Unified Multimodal Models via Decompositional Verifiable Reward
Runhui Huang, Jie Wu, Rui Yang, Zhe Liu, Hengshuang Zhao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use DVReward instead of VIEScore holistic score (ablation) | BAGEL GRPO · GenEval | overall score: 81.7 -> 85.1 | Improved |
| use DVReward instead of VIEScore (ablation) | SD3.5M FlowGRPO · TIIF-Bench Long | score: 72.9 -> 77.7 | Improved |
| use HPSv3 or UnifiedReward as reward (ablation) | BAGEL GRPO · GenEval | overall score: 84.0 -> 83.4 (HPSv3); 84.0 -> 83.7 (UnifiedReward) | Hurt |
| use confidence score instead of binary score in DVReward (ablation) | AlphaGRPO RT2I · GenEval | overall score: 84.0 -> 85.1 | Improved |
| use confidence score instead of binary score (ablation) | AlphaGRPO RT2I · TIIF-Bench Short | score: 79.0 -> 78.9 | No clear effect |
| add False-Positive Rectification (ablation) | AlphaGRPO self-reflective refinement · TIIF-Bench Long | score: 77.8 -> 79.5 | Improved |
| add perceptual quality rubrics to semantic rubrics (ablation) | DVReward · TIIF-Bench Long (1024) | score: 77.7 -> 79.5 | Improved |
| add perceptual quality rubrics (ablation) | DVReward · GenEval (1024) | overall score: 87.3 -> 87.4 | No clear effect |
| add inference-time self-reflective refinement | AlphaGRPO · TIIF-Bench Overall Short (512) | score: 79.1 -> 83.9 | Improved |
| train on self-reflective refinement (no editing data) | AlphaGRPO · GEdit-Bench-EN overall | score: 6.52 -> 7.08 | Improved |
| train on reasoning T2I instead of self-reflective refinement | AlphaGRPO · GEdit-Bench-EN overall | score: 6.52 -> 6.85 | Improved |
| train at 512 resolution, evaluate at 1024 | AlphaGRPO · GenEval 1024 | score: 86.6 -> 86.1 (AlphaGRPO); 86.6 -> 87.4 (RT2I) | Mixed |
Closest papers by meaning
- Diving into Self-Evolving Training for Multimodal Reasoning (ICML 2025 · similarity 0.89)
- Learning to Generate via Understanding: Understanding-Driven Intrinsic Rewarding for Unified Multimodal Models (CVPR 2026 · similarity 0.89)
- Flex-Judge: Text-Only Reasoning Unleashes Zero-Shot Multimodal Evaluators (NeurIPS 2025 · similarity 0.89)
- SRPO: Enhancing Multimodal LLM Reasoning via Reflection-Aware Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- Towards Unified Multimodal Interleaved Generation via Group Relative Policy Optimization (NeurIPS 2025 · similarity 0.89)
- MindOmni: Unleashing Reasoning Generation in Vision Language Models with RGPO (NeurIPS 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
