Read It Back: Pretrained MLLMs Are Zero-Shot Reward Models for Text-to-Image Generation
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.
Tom: I'm Tom.
Jane: And I'm Jane, here with Lu, Meng and our in-house large language model, Lalam.
Tom: We're looking at the paper: "Read It Back".
Jane: Here's the gist of the paper. In this paper, we propose SpectraReward, a training-free reward function that turns pretrained MLLMs into off-the-shelf reward models for image-generation reinforcement learning because it measures how well the original prompt can be recovered from the generated image through a single imageconditioned, teacher-forced forward pass.
Tom: SpectraReward Mechanism SpectraReward is a training-free reward function that turns any pretrained MLLM into an image-generation reward model by reusing its image-text alignment knowledge.
Jane: Given a generated image, SpectraReward conditions a frozen MLLM on the image and performs a single teacher-forced forward pass on the prompt tokens x, producing a token-level likelihood profile that reflects how well each textual requirement can be read from the image.
Tom: The final reward function is defined as RM(x, y) = 1/T sum from t=1 to T-1 log pM(xt+1 x≤t, y).
Title and authors: Jane: Now let's look at the title again, "Read It Back: Pretrained MLLMs Are Zero-Shot Reward Models for Text-to-Image Generation". What does that really tell us about what this paper is proposing?
Lu: The title suggests a few things. First, it says you can read back what the model understands from an image and prompt. Second, it claims these pre-trained models can act as reward models without needing any training on preferences or fine-tuning. Third, it points directly at text to image generation as the specific task they are focused on.
Tom: I think it’s saying we don't need a separate step to create a reward function; the MLLM already knows how to measure alignment between text and image through its pre-training, so you can just use that.
Meng: So it’s zero-shot in terms of preference labels, which is huge for efficiency, but what about the actual quality of that reward? Is it reliable enough for real deployment?
Jane: That’s the question. The paper proposes using the average image conditioned prompt log likelihood as this reward, and they show it works across a wide range of models and benchmarks. They are essentially reusing whatever image-text alignment ability you already have in your pre-trained model to judge generation quality.
Lu: They provide quantitative comparisons showing that SpectraReward outperforms prior MLLM derived reward training methods on benchmarks like TIIFBench and GenEval, with gains ranging from six point three to two point one gains on those specific tests.
Tom: Those numbers are encouraging because they show a solid improvement over the previous methods they compared against, but I always want to see how that scales as you go bigger.
Meng: That’s what I’m curious about. If we use a massive model like the two hundred thirty-five billion parameter one, does that performance keep climbing or does it plateau?
Jane: The paper actually shows it climbs pretty well up to the largest models they tested, but there's a point where things start leveling off, which leads us into what they suggest next.
Lu: And that plateau is why they introduce Self-SpectraReward; if you use a unified multimodal model, the policy’s own understanding branch becomes the reward model for its generation branch.
The paper's summary: Tom: So, to summarize what they’re actually doing, SpectraReward is a training-free reward function that turns any pre-trained MLLM into an off-the-shelf reward model for text-to-image reinforcement learning.
Jane: That means instead of spending time collecting human preference labels or fine-tuning a reward model, you condition the frozen MLLM on the generated image and prompt to get a token likelihood profile that reflects how well each textual requirement can be read from the image.
Lu: The final reward is calculated as RM of x y equals one over T sum from t equals one to T minus one log p M of x t plus one x less than or equal to t, y, which is the mean image conditioned prompt log likelihood under M and those token-wise likelihoods form the semantic spectrum of the image-text pair.
Meng: So they are essentially measuring how well each part of your prompt can be read from what’s in the image, and that score becomes your reward signal for RL training. That seems like a very direct way to link text fidelity to generation quality.
Tom: It is direct because it doesn't require you to ask the MLLM to judge or answer decomposed verification questions, which cuts out those extra steps.
Jane: But what about the limitations they mention? They admit that the reward quality depends entirely on the visual reasoning ability of whatever pre-trained MLLM you pick as your backbone.
Lu: That’s true, and they also point out that because this reward is computed only over the input prompt, it might under-emphasize implicit visual implications that aren't directly in the text.
The paper's improvements: Tom: So the main improvements they highlight are two things: SpectraReward itself and then Self-SpectraReward, which is their special case for unified multimodal models.
Jane: The improvement with Self-SpectraReward is that it forms a closed-loop self-improving framework without any external reward models or external knowledge whatsoever. This means the model improves its own generation quality by using its own understanding ability to improve generation quality.
Lu: That’s the core idea, and they show that this can match or even surpass much larger external reward models when the policy and the reward model are well-aligned. Distribution between them is key here.
Meng: So if I'm deploying a model, am I relying on this alignment being perfect for that to be a reliable way to go? What’s the risk if that alignment isn't perfect?
Tom: The risk is that the paper found it under-emphasizes things like aesthetics and safety, so you still need to add those specialized filters.
Jane: So while Self-SpectraReward shows promise for self-improvement, SpectraReward is a great tool because it’s training-free and off-the-shelf for any pre-trained model you have lying around.
Lu: That’s the practical takeaway: reward function and policy alignment can be just as important as the scale of the reward model; a well aligned self-reward can match or surpass much larger external models.
Conclusion: Tom: So, wrapping up, the authors introduced SpectraReward and Self-SpectraReward, showing that you can turn pre-trained MLLMs into zero-shot reward models for text to image generation.
Jane: This means for you're not just collecting preference labels or fine-tuning a reward model anymore; you can get high quality rewards directly from the model’s inherent alignment knowledge.
Lu: The key takeaway is that the paper is showing that distributional alignment between the reward model and policy is a key factor in effective reinforcement learning in unified multimodal generation.
Meng: So, for production, this tool seems like it simplifies things by providing a training-free path to better performance across different models.
Tom: It does simplify the process of getting those initial RL steps started without needing any external reward model setup.
Jane: And they’re moving towards a future where models can use their own understanding to improve generation quality autonomously.
Lu: That self-improving framework is a really ambitious direction for unified multimodal architectures.
Runhui Huang, Qihui Zhang, Zhe Liu, Yu Gao, Jie Wu, Hengshuang Zhao
University of Hong Kong · ByteDance Seed · Peking University
cs.CV
Submitted: 2026-07-13
Updated: 2026-11-06
Code: https://github.com/black-forest-labs/flux
Project page: https://huangrh99.github.io/SpectraReward
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 100/100
The gist: In this paper, we propose SpectraReward, a training-free reward function that turns pretrained MLLMs into off-the-shelf reward models for image-generation reinforcement learning because it measures…
Key concepts
- SpectraReward Mechanism
- This is a training-free method that uses a frozen MLLM to evaluate generated images. It conditions the MLLM on an image and checks how well the original prompt tokens can be recovered from it. The reward is calculated based on the token-level likelihood profile, which reflects how well each textual requirement is satisfied by the image.
- Self-SpectraReward Framework
- This framework applies SpectraReward to unified multimodal models where the policy's own understanding branch acts as the reward model. This creates a self-improving loop without needing external reward models. The model uses its internal visual understanding to directly provide reward signals for its generation branch, improving alignment.
- Token-level Likelihood Profile
- This is a set of scores generated by the MLLM after processing an image and a prompt. Each token in the prompt gets a likelihood score indicating how likely that specific text element is to be present or correctly represented in the generated image. These scores form the 'semantic spectrum' of the image-text pair.
- Reward-Policy Alignment
- This refers to how well a model's reward signal matches its generation policy. The paper shows that aligning these two components—the reward function and the policy—is key for effective reinforcement learning in image generation. A well-aligned self-reward can outperform much larger external reward models.
Terminology
Summary
In this paper, we propose SpectraReward, a training-free reward function that turns pretrained MLLMs into off-the-shelf reward models for image-generation reinforcement learning because it measures how well the original prompt can be recovered from the generated image through a single imageconditioned, teacher-forced forward pass. p. 0
SpectraReward Mechanism
SpectraReward is a training-free reward function that turns any pretrained MLLM into an image-generation reward model by reusing its image-text alignment knowledge. Given a generated image, SpectraReward conditions a frozen MLLM on the image and performs a single teacher-forced forward pass on the prompt tokens x, producing a token-level likelihood profile that reflects how well each textual requirement can be read from the image. p. 1 The final reward function is defined as RM(x, y) = 1/T sum from t=1 to T-1 log pM(xt+1 x≤t, y). p. 3 This reward is the mean image-conditioned prompt log-likelihood under M and the token-wise likelihoods form the semantic spectrum of the image-text pair. p. 3
Self-SpectraReward Framework
The paper introduces Self-SpectraReward, a special case for unified multimodal models where the policy’s own understanding branch serves as the reward model for its generation branch. This forms a closed-loop self-improving framework without external reward models or external knowledge. p. 2 Since a UMM integrates image understanding and image generation into a single framework, its understanding branch could directly provide the reward signals for its generation branch. p. 1 This design eliminates the need for an external model and improves reward-policy alignment. p. 3
Experimental Validation
Extensive experiments validate SpectraReward through a broad image-generation RL study covering two diffusion models, three RL algorithms, nine reward MLLM backbones from four MLLM families spanning 4B to 235B parameters, and five out-of-distribution text-toimage benchmarks. Results show that both SpectraReward and Self-SpectraReward significantly and consistently improve generation performance and outperform prior MLLM-derived reward training methods. p. 0 Larger reward MLLMs are not always better, while Self-SpectraReward can match or surpass much larger external reward models, suggesting that reward-policy alignment is a key factor for effective image-generation RL. p. 0
Key Findings and Analysis
The analysis reveals several important properties of the reward signal. The token-level likelihoods are sensitive to targeted semantic errors, as seen when comparing positive and negative image pairs for attribute mismatch and object identity mismatch. p. 4 SpectraReward assigns higher rewards to images that better satisfy the prompt and lower rewards to images with missing objects, wrong attributes, or incorrect spatial relations. p. 5 Furthermore, Self-SpectraReward exhibits better improvements compared to SpectraReward with similar-scale MLLM backbones, and comparable or even better performance compared to 4x or 30x larger MLLMs. p. 2 This suggests that distributional alignment between the reward model and the policy can be as important as the scale of the reward model. p. 2
Limitations and Future Directions
SpectraReward relies on the image-conditioned prompt likelihood of pretrained MLLMs, so its reward quality is bounded by the visual understanding and reasoning ability of the chosen reward backbone. Since the likelihood is computed only over the input prompt, it may under-emphasize implicit visual implications that are not directly expressed in the text. p. 4 SpectraReward does not directly optimize criteria such as aesthetics and safety, which could be addressed by combining it with complementary specialized reward models. p. 0 p. 0 The paper also notes that the EOS token is excluded from the computation because its likelihood is often dominated by sequence termination rather than image-text alignment. p. 2 p. 2
Additional Implementation Details
When using SpectraReward, we do not add any general prefix to ask the reward MLLM to describe the image by default. The only exception is the InternVL3. p. 1 p. 15 family, where adding the prefix “Describe the image” gives more stable reward values. p. 8 p. 8 For Self-SpectraReward on BAGEL, we include the VAE feature in Self-SpectraReward because it gives better average performance in our ablation. p. 0 The results show that including the VAE feature obtains better performance on average, and the improvement is more significant on the TIIF-Long benchmark. p. 5 p. 5 The sequence-level reward is used by default for more stable training. p. 8
Broader Impacts
SpectraReward lowers the cost of constructing reward signals for image-generation RL by reusing frozen pretrained MLLMs, without preference annotation or reward-model fine-tuning. This may make reinforcement learning for text-to-image generation more accessible to researchers with limited annotation or training resources. p. 0 However, easier RL optimization for image generation may amplify the existing risks of text-to-image models. p. 4 Stronger prompt following can be misused for generating misleading visual content, imitating protected styles, or producing unsafe images. p. 3 We therefore recommend using SpectraReward together with standard safety filters, bias auditing, and human evaluation when deploying or releasing models trained with this reward. p. 2 The paper also hypothesizes that larger MLLMs may devote more capacity to advanced vision-language reasoning and instruction-following tasks. p. 7 The pretraining-stage MLLMs can be better reward backbones for SpectraReward because they learn large-scale image-captioning and image-text alignment objectives that closely match the reward function. p. 7
Conclusion
We introduced SpectraReward, a training-free reward function that turns pretrained MLLMs into off-the-shelf reward models for image-generation reinforcement learning by measuring imageconditioned prompt likelihood. We further introduced Self-SpectraReward for unified multimodal models, where the policy’s own understanding branch serves as the reward model for its generation branch. p. 8 Across generator backbones, RL algorithms, reward MLLM backbones, and out-of-distribution benchmarks, SpectraReward significantly and consistently improves text-to-image generation and outperforms prior MLLM-derived reward training methods. p. 8 The results also show that larger reward MLLMs are not always better, while a well-aligned self-reward can match or surpass much larger external reward models. p. 8 These findings suggest that reward function and reward-policy alignment are key factors for reinforcement learning in unified multimodal generation. p. 8
References
[1] Ying Ba, Tianyu Zhang, Yalong Bai, Wenyi Mo, Tao Liang, Bing Su, and Ji-Rong Wen. Enhancing reward models for high-quality image generation: Beyond text-image alignment. p. 9 In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 19022–19031, 2025
[4] Kevin Black, Michael Janner, Yilun Du, Ilya Kostrikov, and Sergey Levine. Training diffusion models with reinforcement learning. p. 9 arXiv preprint arXiv:2305. p. 913301
[9] Chaorui Deng, Deyao Zhu, Kunchang Li, Chenhui Gou, Feng Li, Zeyu Wang, Shu Zhong, Weihao Yu, Xiaonan Nie, Ziang Song, Guang Shi, and Haoqi Fan. Emerging properties in unified multimodal pretraining
[13] Dhruba Ghosh, Hannaneh Hajishirzi, and Ludwig Schmidt. Geneval: an object-focused framework for evaluating text-to-image alignment
[28] Jie Liu, Gongye Liu, Jiajun Liang, Yangguang Li, Jiaheng Liu, Xintao Wang, Pengfei Wan, Di Zhang, and Wanli Ouyang. Flow-grpo: Training flow matching models via online rl
[31] Yuwei Niu, Munan Ning, Mengren Zheng, Weiyang Jin, Bin Lin, Peng Jin, Jiaqi Liao, Kunpeng Ning, Chaoran Feng, Bin Zhu, and Li Yuan. Wise: A world knowledge-informed semantic evaluation for text-to-image generation
[33] Jiadong Pan, Liang Li, Yuxin Peng, Yu-Ming Tang, Shuohuan Wang, Yu Sun, Hua Wu, Qingming Huang, and Haifeng Wang. Learning to generate via understanding: Understanding-driven intrinsic rewarding for unified multimodal models
[37] Sheila Schoepp, Masoud Jafaripour, Yingyue Cao, Tianpei Yang, Fatemeh Abdollahi, Shadan Golestan, Zahin Sufiyan, Osmar R Zaiane, and Matthew E Taylor. The evolving landscape of llm-and vlm-integrated reinforcement learning
[40] Wentao Tan, Qiong Cao, Yibing Zhan, Chao Xue, and Changxing Ding. Beyond human data: Aligning multimodal large language models by iterative self-evolution
[59] Jiazheng Xu, Xiao Liu, Yuchen Wu, Yuxuan Tong, Qinkai Li, Ming Ding, Jie Tang, and Yuxiao Dong. Imagereward: Learning and evaluating human preferences for text-to-image generation
[61] Shuchen Xue, Chongjian Ge, Shilong Zhang, Yichen Li, and Zhi-Ming Ma. Advantage weighted matching: Aligning rl with pretraining in diffusion models
[63] Weizhe Yuan, Richard Yuanzhe Pang, Kyunghyun Cho, Xian Li, Sainbayar Sukhbaatar, Jing Xu, and Jason Weston. Self-rewarding language models
[58] Jinheng Xie, Weijia Mao, Zechen Bai, David Junhao Zhang, Weihao Wang, Kevin Qinghong Lin, Yuchao Gu, Zhijie Chen, Zhenheng Yang, and Mike Zheng Shou. Show-o: One single transformer to unify multimodal understanding and generation
[57] Jinheng Xie, Weijia Mao, Zechen Bai, David Junhao Zhang, Weihao Wang, Kevin Qinghong Lin, Yuchao Gu, Zhijie Chen, Zhenheng Yang, and Mike Zheng Shou. Show-o: One single transformer to unify multimodal understanding and generation
[51] Jie Wu, Yu Gao, Zilyu Ye, Ming Li, Liang Li, Hanzhong Guo, Jie Liu, Zeyue Xue, Xiaoxia Hou, Wei Liu, et al. Rewarddance: Reward scaling in visual generation
[53] Tianhao Wu, Weizhe Yuan, Olga Golovneva, Jing Xu, Yuandong Tian, Jiantao Jiao, Jason E Weston and Sainbayar Sukhbaatar. Meta-rewarding language models: Self-improving alignment with llm-as-a-meta-judge
[55] Yecheng Wu, Zhuoyang Zhang, Junyu Chen, Haotian Tang, Dacheng Li, Yunhao Fang, Ligeng Zhu, Enze Xie, Hongxu Yin, Li Yi, Song Han and Yao Lu. Vila-u: A unified foundation model integrating visual understanding and generation
[56] Shitao Xiao, Yueze Wang, Junjie Zhou, Huaying Yuan, Xingrun Xing, Ruiran Yan, Chaofan Li, Shuting Wang, Tiejun Huang, and Zheng Liu. Omnigen: Unified image generation
[59] Jiazheng Xu, Xiao Liu, Yuchen Wu, Yuxuan Tong, Qinkai Li, Ming Ding, Jie Tang, and Yuxiao Dong. Imagereward: Learning and evaluating human preferences for text-to-image generation
[60] Jiazheng Xu, Yu Huang, Jiale Cheng, Yuanming Yang, Jiajun Xu, Yuan Wang, Wenbo Duan, Shen Yang, Qunlin Jin, Shurun Li et al. Visionreward: Fine-grained multi-dimensional human preference learning for image and video generation
[65] Kaiwen Zheng, Huayu Chen, Haotian Ye, Haoxiang Wang, Qinsheng Zhang, Kai Jiang, Hang Su, Stefano Ermon, Jun Zhu and Ming-Yu Liu. Diffusionnft: Online diffusion reinforcement with forward process
[66] Chunting Zhou, LILI YU, Arun Babu, Kushal Tirumala, Michihiro Yasunaga, Leonid Shamis, Jacob Kahn, Xuezhe Ma, Luke Zettlemoyer and Omer Levy. Transfusion: Predict the next token and diffuse images with one multi-modal model
[62] Zeyue Xue, Jie Wu, Yu Gao, Fangyuan Kong, Lingting Zhu, Mengzhao Chen, Zhiheng Liu, Wei Liu, Qiushan Guo and Weilin Huang. Dancegrpo: Unleashing grpo on visual generation
[64] Shanshan Zhao, Xinjie Zhang, Jintao Guo, Jiakui Hu, Lunhao Duan, Minghao Fu, Yong Xien Chng, Guo-Hua Wang, Qing-Guo Chen, Zhao Xu et al. Unified multimodal understanding and generation models: Advances challenges and opportunities
[67] Kaiwen Zheng, Huayu Chen, Haotian Ye, Haoxiang Wang, Qinsheng Zhang, Kai Jiang, Hang Su, Stefano Ermon, Jun Zhu and Ming-Yu Liu. Diffusionnft: Online diffusion reinforcement with forward process
[8] Yufeng Cui, Honghao Chen, Haoge Deng, Xu Huang, Xinghang Li, Jirong Liu, Yang Liu, Zhuoyan Luo, Jinsheng Wang, Wenxuan Wang et al. Emu3. 5: Native multimodal models are world learners
[10] Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel et al. Scaling rectified flow transformers for high-resolution image synthesis
[12] Yuying Ge, Sijie Zhao, Jinguo Zhu, Yixiao Ge, Kun Yi, Lin Song, Chen Li, Xiaohan Ding and Ying Shan. Seed-x: Multimodal models with unified multi-granularity comprehension and generation
[17] Runhui Huang, Jie Wu, Rui Yang, Zhe Liu and Hengshuang Zhao. Alphagrpo: Unlocking selfreflective multimodal generation in unified multimodal models via decompositional verifiable reward
[38] Zhihong Shao et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models
[43] Gemma Team. Gemma 3 technical report, 2025
[45] Weiyun Wang, Zhangwei Gao, Lixin Gu, Hengjun Pu, Long Cui, Xingguang Wei, Zhaoyang Liu, Linglin Jing, Shenglong Ye et al. Internvl3. 5: Advancing open-source multimodal models in versatility reasoning and efficiency
[47] Yibin Wang, Yuhang Zang, Hao Li, Cheng Jin and Jiaqi Wang. Unified reward model for multimodal understanding and generation
[48] Xinyu Wei, Jinrui Zhang, Zeqing Wang, Hongyang Wei, Zhen Guo and Lei Zhang. Tiif-bench: How does your t2i model follow your instructions? p. 1 arXiv preprint arXiv:2506. p. 002161
[54] Xiaoshi Wu, Yiming Hao, Keqiang Sun, Yixiong Chen, Feng Zhu, Rui Zhao and Hongsheng Li. Human preference score v2: A solid benchmark for evaluating human preferences of text-to-image synthesis
[57] Jinheng Xie, Weijia Mao, Zechen Bai, David Junhao Zhang, Weihao Wang, Kevin Qinghong Lin, Yuchao Gu, Zhijie Chen, Zhenheng Yang and Mike Zheng Shou. Show-o: One single transformer to unify multimodal understanding and generation
[60] Jiazheng Xu, Yu Huang, Jiale Cheng, Yuanming Yang, Jiajun Xu et al. Visionreward: Fine-grained multi-dimensional human preference learning for image and video generation
[52] Shengqiong Wu, Hao Fei, Leigang Qu, Wei Ji and Tat-Seng Chua. Next-gpt: Any-to-any multimodal llm
[39] Zhiqing Sun, Yikang Shen, Hongxin Zhang, Qinhong Zhou, Zhenfang Chen, David Cox, Yiming Yang and Chuang Gan. Salmon: Self-alignment with instructable reward models
[41] Chameleon Team. Chameleon: Mixed-modal early-fusion foundation models
[42] Chameleon Team. Chameleon: Mixed-modal early-fusion foundation models
Appendix
A Limitations and future work discusses the limitations of prompt-likelihood rewards, including dependence on MLLM visual reasoning, implicit physical and commonsense implications, and complementary reward signals. SpectraReward does not directly optimize criteria such as aesthetics and safety, which could be addressed by combining it with complementary specialized reward models. The paper also notes that the EOS token is excluded from the computation because its likelihood is often dominated by sequence termination rather than image-text alignment.
When using SpectraReward, we do not add any general prefix to ask the reward MLLM to describe the image by default. The only exception is the InternVL3.5 family, where adding the prefix “Describe the image” gives more stable reward values. We exclude the [EOS] token from the SpectraReward computation and use both ViT features from the semantic encoder and VAE features from the generation encoder in Self-SpectraReward on BAGEL because it gives better average performance in our ablation. p. 3 The results show that including the VAE feature obtains better performance on average, and the improvement is more significant on the TIIF-Long benchmark.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Reward Function SpectraReward can be used as a training-free reward function by
measuring how well the original prompt can be recovered from the generated image through a single imageconditioned, teacher-forced forward pass.
This allows for obtaining adeterministic, off-the-shelf reward signal
without requiringpreference labels, reward-model fine-tuning, scalar judging, or question decomposition.
-
SelfSpectraReward enables unified multimodal models to form a
closed-loop selfimproving framework without external reward models or external knowledge.
This allows the model to improve its own generation byusing its own understanding ability to improve generation quality,
leveraging the fact thatthe reward signal is compatible with the policy’s own generation knowledge.
-
The system can achieve better performance on complex, fine-grained instruction following and composition by using SpectraReward, as it shows improvements in
counting, spatial position and color-attribute binding
which require not only object presence but alsocorrect attribute binding to objects.
-
SelfSpectraReward can match or surpass much larger external reward models by demonstrating that
reward-policy alignment is a key factor for effective image-generation RL,
suggesting that the model's own understanding branch provides a reward signal that isbetter calibrated to the policy’s own generated-image distribution.
Sources
- Qwen3-VL Technical Report
- Training Diffusion Models with Reinforcement Learning
- BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset
- Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling
- Emu3.5: Native Multimodal Models are World Learners
- Emerging Properties in Unified Multimodal Pretraining
- HonestLLM: Toward an Honest and Helpful Large Language Model
- SEED-X: Multimodal Models with Unified Multi-granularity Comprehension and Generation
- OneReward: Unified Mask-Guided Image Generation via Multi-Task Human Preference Learning
- ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment
- ILLUME+: Illuminating Unified MLLM with Dual Visual Tokenization and Diffusion Refinement
- Modality-Fair Preference Optimization for Trustworthy MLLM Alignment
- SRUM: Fine-Grained Self-Rewarding for Unified Multimodal Models
- GenEval 2: Addressing Benchmark Drift in Text-to-Image Evaluation
- Hunyuan-DiT: A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding
- Uniworld-V2: Reinforce Image Editing with Diffusion Negative-aware Finetuning and MLLM Implicit Feedback
- UniWorld-V1: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation
- Flow-GRPO: Training Flow Matching Models via Online RL
- PromptEcho: Annotation-Free Reward from Vision-Language Models for Text-to-Image Reinforcement Learning
- WISE: A World Knowledge-Informed Semantic Evaluation for Text-to-Image Generation
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs