AutoRubric-T2I: Robust Rule-Based Reward Model for Text-to-Image Alignment

Kuei-Chun Kao, Daixuan Huo, Yuanhao Ban, Cho-Jui Hsieh

arXiv:2605.17602 · 2026-05-23 공개 · arXiv · PDF

diffusion-models vlm text-to-image preference-learning reward-model rubric-learning llr-refiner mmrb2

Abstract

Aligning Text-to-Image (T2I) generation models with human preferences increasingly relies on image reward models that score or rank generated images according to prompt alignment and perceptual quality. Existing reward models are commonly trained as Bradley-Terry (BT) preference models on large-scale human preference corpora, making them costly to train, difficult to adapt, and opaque in their evaluation criteria. Meanwhile, Vision-Language Model (VLM) judges can provide more fine-grained assessments through textual rubrics, but their manually designed or heuristically generated scoring rules may fail to reliably reflect human preferences. In this paper, we propose AutoRubric-T2I, the first rubric learning framework in T2I that automatically synthesizes and selects explicit rubrics for guiding VLM judges. AutoRubric-T2I first synthesizes reasoning traces from preference pairs into candidate rubrics, then uses a VLM judge to score paired images under each rubric, producing pairwise rubric-score differences for preference learning. To remove noisy and redundant rules, we further employ a ell_1-Regularized Logistic Regression Refiner, which selects the Top-N most discriminative rubrics. Extensive evaluations show that AutoRubric-T2I produces high-quality, interpretable reward signals using less than 0.01% of the annotated preference data, substantially reducing the need for large-scale reward-model training. On image reward benchmarks such as MMRB2, AutoRubric-T2I outperforms strong reward model baselines. We further validate AutoRubric-T2I as an RL reward on downstream T2I tasks, including TIIF and UniGenBench++, where it improves generation quality over scalar reward models using the Flow-GRPO pipeline on diffusion models.

한국어 요약

📋 한 줄 요약

**[T2I Reward Model / Auto Rubrics]** AutoRubric-T2I가 preference pair에서 reasoning trace를 자동 합성해 후보 rubric 생성·ℓ1 logistic regression으로 Top-N 선택, 0.01% 미만 annotated data로 강한 reward model baseline 능가·Flow-GRPO 다운스트림에서 생성 품질 향상.

🎯 핵심 기여도

💡 핵심 아이디어

T2I reward model은 대규모 preference annotation 없이 preference pair의 reasoning trace로부터 rubric 자체를 자동 합성하고 ℓ1 sparsity로 가장 discriminative한 Top-N을 선택하면, 최소 데이터로 해석 가능·고품질 reward 신호를 만들 수 있다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: 대규모 preference annotation 의존 탈피, rubric 기반의 해석 가능 reward 신호 자동화, 다운스트림 RL에서도 실효 검증. **한계**: VLM judge 품질에 강하게 의존, rubric 합성 reasoning trace 자체의 신뢰성 문제, T2I 도메인 특화 — video·3D 등 일반화는 후속 과제.

🚀 실용적 활용