# RationalRewards: Reasoning Rewards Scale Visual Generation Both Training and Test Time

arXiv: https://arxiv.org/abs/2604.11626 · HTML: https://mlsanity.com/paper/2604.11626/

## Key facts

- Cited by 14 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion and multimodal model alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 45 of its 53 references that are in the map.
- First posted on arXiv 13 Apr 2026.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Diffusion and multimodal model alignment](https://mlsanity.com/alignment/diffusion-and-multimodal-model-alignment/)

## Builds on (most cited first, 30 of 45)

- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis](https://mlsanity.com/paper/2307.01952/) (2023)
- [Scaling Rectified Flow Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2403.03206/) (2024)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters](https://mlsanity.com/paper/2408.03314/) (2024)
- [ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation](https://mlsanity.com/paper/2304.05977/) (2023)
- [Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation](https://mlsanity.com/paper/2305.01569/) (2023)
- [STaR: Bootstrapping Reasoning With Reasoning](https://mlsanity.com/paper/2203.14465/) (2022)
- [Emerging Properties in Unified Multimodal Pretraining](https://mlsanity.com/paper/2505.14683/) (2025)
- [Flow-GRPO: Training Flow Matching Models via Online RL](https://mlsanity.com/paper/2505.05470/) (2025)
- [Step1X-Edit: A Practical Framework for General Image Editing](https://mlsanity.com/paper/2504.17761/) (2025)
- [DanceGRPO: Unleashing GRPO on Visual Generation](https://mlsanity.com/paper/2505.07818/) (2025)
- [Emu Edit: Precise Image Editing via Recognition and Generation Tasks](https://mlsanity.com/paper/2311.10089/) (2023)
- [VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning](https://mlsanity.com/paper/2504.08837/) (2025)
- [Improving Video Generation with Human Feedback](https://mlsanity.com/paper/2501.13918/) (2025)
- [Pixel Reasoner: Incentivizing Pixel-Space Reasoning with Curiosity-Driven Reinforcement Learning](https://mlsanity.com/paper/2505.15966/) (2025)
- [UltraEdit: Instruction-based Fine-Grained Image Editing at Scale](https://mlsanity.com/paper/2407.05282/) (2024)
- [OmniGen2: Towards Instruction-Aligned Multimodal Generation](https://mlsanity.com/paper/2506.18871/) (2025)
- [AnyEdit: Mastering Unified High-Quality Image Editing for Any Idea](https://mlsanity.com/paper/2411.15738/) (2024)
- [DiffusionNFT: Online Diffusion Reinforcement with Forward Process](https://mlsanity.com/paper/2509.16117/) (2025)
- [HPSv3: Towards Wide-Spectrum Human Preference Score](https://mlsanity.com/paper/2508.03789/) (2025)
- [Unified Reward Model for Multimodal Understanding and Generation](https://mlsanity.com/paper/2503.05236/) (2025)
- [T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT](https://mlsanity.com/paper/2505.00703/) (2025)
- [HunyuanVideo 1.5 Technical Report](https://mlsanity.com/paper/2511.18870/) (2025)
- [Generative Reward Models](https://mlsanity.com/paper/2410.12832/) (2024)
- [OmniEdit: Building Image Editing Generalist Models Through Specialist Supervision](https://mlsanity.com/paper/2411.07199/) (2024)
- [GenAI Arena: An Open Evaluation Platform for Generative Models](https://mlsanity.com/paper/2406.04485/) (2024)
- [Uniworld-V2: Reinforce Image Editing with Diffusion Negative-aware Finetuning and MLLM Implicit Feedback](https://mlsanity.com/paper/2510.16888/) (2025)
- [Ovis-U1 Technical Report](https://mlsanity.com/paper/2506.23044/) (2025)

## Built on by (most cited first, 14 of 14)

- [Search Beyond What Can Be Taught: Evolving the Knowledge Boundary in Agentic Visual Generation](https://mlsanity.com/paper/2607.05382/) (2026)
- [Bad Seeing or Bad Thinking? Rewarding Perception for Multimodal Reasoning](https://mlsanity.com/paper/2605.14054/) (2026)
- [SenseNova-U1.5: Towards Native Unified Visual Intelligence](https://mlsanity.com/paper/2609.11929/) (2026)
- [PanoWorld: Towards Spatial Supersensing in 360^circ Panorama World](https://mlsanity.com/paper/2605.13169/) (2026)
- [Rethinking Classifier-Free Guidance in On-Policy Diffusion Distillation](https://mlsanity.com/paper/2607.24731/) (2026)
- [G2TR: Generation-Guided Visual Token Reduction for Separate-Encoder Unified Multimodal Models](https://mlsanity.com/paper/2605.12309/) (2026)
- [Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing](https://mlsanity.com/paper/2607.19064/) (2026)
- [AVA-Encoder: Towards Agent-Native Video Representation Learning](https://mlsanity.com/paper/2608.12313/) (2026)
- [SpecPath: Testing Coding Agents Across Contract-Equivalent Specification Histories](https://mlsanity.com/paper/2608.09799/) (2026)
- [Starve to Perceive: Taming Lazy Perception in VLMs with Constrained Visual Bandwidth](https://mlsanity.com/paper/2605.18603/) (2026)
- [The Missing Temporal Link: Temporal Context Routing for Script-Driven Audio-Video Generation](https://mlsanity.com/paper/2609.02367/) (2026)
- [SciGen-Verifier: A Multimodal Reasoner for Explainable Verification in Scientific Image Generation](https://mlsanity.com/paper/2609.33399/) (2026)
- [Think Before You Score: Thinking Reward Model for Visual Generation](https://mlsanity.com/paper/2609.37372/) (2026)
- [Getting Space Right: Agentic Rewards for Image Generation](https://mlsanity.com/paper/2610.14852/) (2026)

## Abstract (arXiv)

Most reward models for visual generation reduce rich human judgments to a single unexplained score, discarding the reasoning that underlies preference. We show that teaching reward models to produce explicit, multi-dimensional critiques before scoring transforms them from passive evaluators into active optimization tools, improving generators in two complementary ways: at training time, structured rationales provide interpretable, fine-grained rewards for reinforcement learning; at test time, a Generate-Critique-Refine loop turns critiques into targeted prompt revisions that improve outputs without any parameter updates. To train such a reward model without costly rationale annotations, we introduce Preference-Anchored Rationalization (PARROT), a principled framework that recovers high-quality rationales from readily available preference data through anchored generation, consistency filtering, and distillation. The resulting model, RationalRewards (8B), achieves state-of-the-art preference prediction among open-source reward models, competitive with Gemini-2.5-Pro, while using 10-20x less training data than comparable baselines. As an RL reward, it consistently improves text-to-image and image-editing generators beyond scalar alternatives. Most strikingly, its test-time critique-and-refine loop matches or exceeds RL-based fine-tuning on several benchmarks, suggesting that structured reasoning can unlock latent capabilities in existing generators that suboptimal prompts fail to elicit.