Post-Training Frontier Text-to-Image Models by Composing Preference and Rubric Rewards
cs.CV, cs.AI
Submitted: 2026-10-02
Updated: 2026-10-02
Code: https://github.com/ideogram-oss/ideogram4
Project page: https://banyuanhao.github.io/PostTraining-Frontier-Text-to-Image-Modelsby
Sources
- Qwen3-VL Technical Report
- Arena-T2I Hard: Benchmarking and Improving Faithfulness with Dependency-Aware Checklist
- Davidsonian Scene Graph: Improving Reliability in Fine-grained Evaluation for Text-to-Image Generation
- GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment
- TempFlow-GRPO: When Timing Matters for GRPO in Flow Models
- CLIPScore: A Reference-free Evaluation Metric for Image Captioning
- TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering
- Multimodal RewardBench 2: Evaluating Omni Reward Models for Interleaved Text and Image
- AutoRubric-T2I: Robust Rule-Based Reward Model for Text-to-Image Alignment
- Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation
- MixGRPO: Unlocking Flow-based GRPO Efficiency with Mixed ODE-SDE
- Flow-GRPO: Training Flow Matching Models via Online RL
- GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization
- SpatialReward: Bridging the Perception Gap in Online RL for Image Editing via Explicit Spatial Reasoning
- HPSv3: Towards Wide-Spectrum Human Preference Score
- Direct Preference Optimization: Your Language Model is Secretly a Reward Model
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria
- Diffusion Model Alignment Using Direct Preference Optimization
- AutoRule: Reasoning Chain-of-thought Extracted Rule-based Rewards Improve Preference Learning
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models