NeurIPS 2025 Papers

Text-to-image generation has evolved beyond single monolithic models to complex multi-component pipelines that combine various enhancement tools. While these pipelines significantly improve image quality, their effective design requires substantial expertise. Recent approaches automating this process through large language models (LLMs) have shown promise but suffer from two critical limitations: extensive computational requirements from generating images with hundreds of predefined pipelines, and poor generalization beyond memorized training examples. We introduce a novel reinforcement learning-based framework that addresses these inefficiencies. Our approach first trains an ensemble of reward models capable of predicting image quality scores directly from prompt-workflow combinations, eliminating the need for costly image generation during training. We then implement a two-phase training strategy: initial workflow prediction training followed by GRPO-based optimization that guides the model toward higher-performing regions of the workflow space. Additionally, we incorporate a classifier-free guidance based enhancement technique that extrapolates along the path between the initial and GRPO-tuned models, further improving output quality. We validate our approach through a set of comparisons, showing that it can successfully create new flows with greater diversity and lead to superior image quality compared to existing baselines.

We introduce a model named DreamLight for universal image relighting in this work, which can seamlessly composite subjects into a new background while maintaining aesthetic uniformity in terms of lighting and color tone. The background can be specified by natural images (image-based relighting) or generated from unlimited text prompts (text-based relighting). Existing studies primarily focus on image-based relighting, while with scant exploration into text-based scenarios. Some works employ intricate disentanglement pipeline designs relying on environment maps to provide relevant information, which grapples with the expensive data cost required for intrinsic decomposition and light source. Other methods take this task as an image translation problem and perform pixel-level transformation with autoencoder architecture. While these methods have achieved decent harmonization effects, they struggle to generate realistic and natural light interaction effects between the foreground and background. To alleviate these challenges, we reorganize the input data into a unified format and leverage the semantic prior provided by the pretrained diffusion model to facilitate the generation of natural results. Moreover, we propose a Position-Guided Light Adapter (PGLA) that condenses light information from different directions in the background into designed light query embeddings, and modulates the foreground with direction-biased masked attention. In addition, we present a post-processing module named Spectral Foreground Fixer (SFF) to adaptively reorganize different frequency components of subject and relighted background, which helps enhance the consistency of foreground appearance. Extensive comparisons and user study demonstrate that our DreamLight achieves remarkable relighting performance.

As a new paradigm of visual content generation, autoregressive text-to-image models suffer from slow inference due to their sequential token-by-token decoding process, often requiring thousands of model forward passes to generate a single image. To address this inefficiency, we propose Speculative Jacobi-Denoising Decoding (SJD2), a framework that incorporates the denoising process into Jacobi iterations to enable parallel token generation in autoregressive models. Our method introduces a next-clean-token prediction paradigm that enables the pre-trained autoregressive models to accept noise-perturbed token embeddings and predict the next clean tokens through low-cost fine-tuning. This denoising paradigm guides the model towards more stable Jacobi trajectories. During inference, our method initializes token sequences with Gaussian noise and performs iterative next-clean-token-prediction in the embedding space. We employ a probabilistic criterion to verify and accept multiple tokens in parallel, and refine the unaccepted tokens for the next iteration with the denoising trajectory. Experiments show that our method can accelerate generation by reducing model forward passes while maintaining the visual quality of generated images.

Recent advancements in video generation have enabled the creation of high-quality, visually compelling videos. However, generating videos that adhere to the laws of physics remains a critical challenge for applications requiring realism and accuracy. In this work, we propose **PhysHPO**, a novel framework for Hierarchical Cross-Modal Direct Preference Optimization, to tackle this challenge by enabling fine-grained preference alignment for physically plausible video generation. PhysHPO optimizes video alignment across four hierarchical granularities: a) ***Instance Level***, aligning the overall video content with the input prompt; b) ***State Level***, ensuring temporal consistency using boundary frames as anchors; c) ***Motion Level***, modeling motion trajectories for realistic dynamics; and d) ***Semantic Level***, maintaining logical consistency between narrative and visuals. Recognizing that real-world videos are the best reflections of physical phenomena, we further introduce an automated data selection pipeline to efficiently identify and utilize *"good data"* from existing large-scale text-video datasets, thereby eliminating the need for costly and time-intensive dataset construction. Extensive experiments on both physics-focused and general capability benchmarks demonstrate that PhysHPO significantly improves physical plausibility and overall video generation quality of advanced models. To the best of our knowledge, this is the first work to explore fine-grained preference alignment and data selection for video generation, paving the way for more realistic and human-preferred video generation paradigms.

Knowledge distillation from Large Language Models (LLMs) to locally hosted Small Language Models (SLMs) provides advantages for Data Analysis Code Generation (DACG) such as privacy protection. However, achieving effective distillation without resource-intensive training is challenging. This paper investigates whether LLMs can distill knowledge to SLMs through In-Context Learning (ICL), a training-free method for rapid task adaptation. We present the DarGO: Distillation and Adaptive Reasoning-Guided Orchestration framework, which facilitates automatic knowledge distillation from LLMs to SLMs. DarGO consists of three phases: exploration through an Model Orchestration Interface (MOI), Memory Collection of successful trajectories, and Knoweldge-driven Inference. We evaluate DarGO on three challenging DACG benchmarks (WikiTQ, TabMWP, and Bird-SQL), each with in-domain training sets that enable detailed analysis of knowledge distillation effectiveness. DarGO demonstrates a substantial relative performance improvement of 27.5\% on average for the student SLMs. To further observe generalization capabilities, we evaluate the \method across different teacher-student model combinations, knowledge transfer scenarios, and unified memory approaches for more advanced, test-only data analysis tasks. Our findings contribute a novel perspective on distillation methods that enhance high performance for SLMs while avoiding intensive fine-tuning.

Text-to-image diffusion models have shown an impressive ability to generate high-quality images from input textual descriptions/prompts. However, concerns have been raised about the potential for these models to create content that infringes on copyrights or depicts disturbing subject matter. Removing specific concepts from these models is a promising solution to this issue. However, existing methods for concept removal do not work well in practical but challenging scenarios where concepts need to be continuously removed. Specifically, these methods lead to poor alignment between the text prompts and the generated image after the continuous removal process. To address this issue, we propose a novel concept removal approach called CCRT that includes a designed knowledge distillation paradigm. CCRT constrains the text-image alignment behavior during the continuous concept removal process by using a set of text prompts. These prompts are generated through our genetic algorithm, which employs a designed fuzzing strategy. To evaluate the effectiveness of CCRT, we conduct extensive experiments involving the removal of various concepts, algorithmic metrics, and human studies. The results demonstrate that CCRT can effectively remove the targeted concepts from the model in a continuous manner while maintaining the high image generation quality (e.g., text-image alignment). The code of CCRT is available at https://github.com/wssun/CCRT.

Directly generating 3D meshes, the default representation for 3D shapes in the graphics industry, using auto-regressive (AR) models has become popular these days, thanks to their sharpness, compactness in the generated results, and ability to represent various types of surfaces. However, AR mesh generative models typically construct meshes face by face in lexicographic order, which does not effectively capture the underlying geometry in a manner consistent with human perception. Inspired by 2D models that progressively refine images, such as the prevailing next-scale prediction AR models, we propose generating meshes auto-regressively in a progressive coarse-to-fine manner. Specifically, we view mesh simplification algorithms, which gradually merge mesh faces to build simpler meshes, as a natural fine-to-coarse process. Therefore, we generalize meshes to simplicial complexes and develop a transformer-based AR model to approximate the reverse process of simplification in the order of level of detail, constructing meshes initially from a single point and gradually adding geometric details through local remeshing, where the topology is not predefined and is alterable. Our experiments show that this novel progressive mesh generation approach not only provides intuitive control over generation quality and time consumption by early stopping the auto-regressive process but also enables applications such as mesh refinement and editing.

We present an ultra-efficient post-training method for shortcutting large-scale pre-trained flow matching diffusion models into efficient few-step samplers, enabled by novel velocity field self-distillation. While shortcutting in flow matching, originally introduced by shortcut models, offers flexible trajectory-skipping capabilities, it requires a specialized step-size embedding incompatible with existing models unless retraining from scratch—a process nearly as costly as pretraining itself. Our key contribution is thus imparting a more aggressive shortcut mechanism to standard flow matching models (e.g., Flux), leveraging a unique distillation principle that obviates the need for step-size embedding. Working on the velocity field rather than sample space and learning rapidly from self-guided distillation in an online manner, our approach trains efficiently, e.g., producing a 3-step Flux <1 A100 day. Beyond distillation, our method can be incorporated into the pretraining stage itself, yielding models that inherently learn efficient, few-step flows without compromising quality. This capability also enables, to our knowledge, the first few-shot distillation method (e.g., 10 text-image pairs) for dozen-billion-parameter diffusion models, delivering state-of-the-art performance at almost free cost.

With the rapid advancement of Artificial Intelligence Generated Content (AIGC) technologies, synthetic images have become increasingly prevalent in everyday life, posing new challenges for authenticity assessment and detection. Despite the effectiveness of existing methods in evaluating image authenticity and locating forgeries, these approaches often lack human interpretability and do not fully address the growing complexity of synthetic data. To tackle these challenges, we introduce FakeVLM, a specialized large multimodal model designed for both general synthetic image and DeepFake detection tasks. FakeVLM not only excels in distinguishing real from fake images but also provides clear, natural language explanations for image artifacts, enhancing interpretability. Additionally, we present FakeClue, a comprehensive dataset containing over 100,000 images across seven categories, annotated with fine-grained artifact clues in natural language. FakeVLM demonstrates performance comparable to expert models while eliminating the need for additional classifiers, making it a robust solution for synthetic data detection. Extensive evaluations across multiple datasets confirm the superiority of FakeVLM in both authenticity classification and artifact explanation tasks, setting a new benchmark for synthetic image detection. The code, model weights, and dataset can be found here: https://github.com/opendatalab/FakeVLM.

Restoring images degraded by adverse weather remains a significant challenge due to the highly non-uniform and spatially heterogeneous nature of weather-induced artifacts, \emph{e.g.}, fine-grained rain streaks versus widespread haze. Accurately estimating the underlying degradation can intuitively provide restoration models with more targeted and effective guidance, enabling adaptive processing strategies. To this end, we propose a Morton-Order Degradation Estimation Mechanism (MODEM) for adverse weather image restoration. Central to MODEM is the Morton-Order 2D-Selective-Scan Module (MOS2D), which integrates Morton-coded spatial ordering with selective state-space models to capture long-range dependencies while preserving local structural coherence. Complementing MOS2D, we introduce a Dual Degradation Estimation Module (DDEM) that disentangles and estimates both global and local degradation priors. These priors dynamically condition the MOS2D modules, facilitating adaptive and context-aware restoration. Extensive experiments and ablation studies demonstrate that MODEM achieves state-of-the-art results across multiple benchmarks and weather types, highlighting its effectiveness in modeling complex degradation dynamics. Our code will be released soon.

Text-to-video generative models convert textual prompts into dynamic visual content, offering wide-ranging applications in film production, gaming, and education. However, their real-world performance often falls short of user expectations. One key reason is that these models have not been trained on videos related to some topics users want to create. In this paper, we propose VideoUFO, the first Video dataset specifically curated to align with Users' FOcus in real-world scenarios. Beyond this, our VideoUFO also features: (1) minimal (0.29\%) overlap with existing video datasets, and (2) videos searched exclusively via YouTube's official API under the Creative Commons license. These two attributes provide future researchers with greater freedom to broaden their training sources. The VideoUFO comprises over 1.09 million video clips, each paired with both a brief and a detailed caption (description). Specifically, through clustering, we first identify 1,291 user-focused topics from the million-scale real text-to-video prompt dataset, VidProM. Then, we use these topics to retrieve videos from YouTube, split the retrieved videos into clips, and generate both brief and detailed captions for each clip. After verifying the clips with specified topics, we are left with about 1.09 million video clips. Our experiments reveal that (1) current 16 text-to-video models do not achieve consistent performance across all user-focused topics; and (2) a simple model trained on VideoUFO outperforms others on worst-performing topics. The dataset and code are publicly available at https://huggingface.co/datasets/WenhaoWang/VideoUFO and https://github.com/WangWenhao0716/BenchUFO under the CC BY 4.0 License.

This paper explores a novel lightweight approach LightFair to achieve fair text-to-image diffusion models (T2I DMs) by addressing the adverse effects of the text encoder. Most existing methods either couple different parts of the diffusion model for full-parameter training or rely on auxiliary networks for correction. They incur heavy training or sampling burden and unsatisfactory performance. Since T2I DMs consist of multiple components, with the text encoder being the most fine-tunable and front-end module, this paper focuses on mitigating bias by fine-tuning text embeddings. To validate feasibility, we observe that the text encoder’s neutral embedding output shows substantial skewness across image embeddings of various attributes in the CLIP space. More importantly, the noise prediction network further amplifies this imbalance. To finetune the text embedding, we propose a collaborative distance-constrained debiasing strategy that balances embedding distances to improve fairness without auxiliary references. However, mitigating bias can compromise the original generation quality. To address this, we introduce a two-stage text-guided sampling strategy to limit when the debiased text encoder intervenes. Extensive experiments demonstrate that LightFair is effective and efficient. Notably, on Stable Diffusion v1.5, our method achieves SOTA debiasing at just $1/4$ of the training burden, with virtually no increase in sampling burden. The code is available at https://github.com/boyuh/LightFair.

Previous methods for image geo-localization have typically treated the task as either classification or retrieval, often relying on black-box decisions that lack interpretability. The rise of large vision-language models (LVLMs) has enabled a rethinking of geo-localization as a reasoning-driven task grounded in visual cues. However, two major challenges persist. On the data side, existing reasoning-focused datasets are primarily based on street-view imagery, offering limited scene diversity and constrained viewpoints. On the modeling side, current approaches predominantly rely on supervised fine-tuning, which yields only marginal improvements in reasoning capabilities. To address these challenges, we propose a novel pipeline that constructs a reasoning-oriented geo-localization dataset, $\textit{MP16-Reason}$, using diverse social media images. We introduce $\textit{GLOBE}$, $\textbf{G}$roup-relative policy optimization for $\textbf{L}$ocalizability assessment and $\textbf{O}$ptimized visual-cue reasoning, yielding $\textbf{B}$i-objective geo-$\textbf{E}$nhancement for the VLM in recognition and reasoning. $\textit{GLOBE}$ incorporates task-specific rewards that jointly enhance localizability assessment, visual-cue reasoning, and geolocation accuracy. Both qualitative and quantitative results demonstrate that $\textit{GLOBE}$ outperforms state-of-the-art open-source LVLMs on geo-localization tasks, particularly in diverse visual scenes, while also generating more insightful and interpretable reasoning trajectories. The data and code are available at https://github.com/lingli1996/GLOBE.

Learning probability models from data is at the heart of many machine learning endeavors, but is notoriously difficult due to the curse of dimensionality. We introduce a new framework for learning \emph{normalized} energy (log probability) models that is inspired from diffusion generative models, which rely on networks optimized to estimate the score. We modify a score network architecture to compute an energy while preserving its inductive biases. The gradient of this energy network with respect to its input image is the score of the learned density, which can be optimized using a denoising objective. Importantly, the gradient with respect to the noise level provides an additional score that can be optimized with a novel secondary objective, ensuring consistent and normalized energies across noise levels. We train an energy network with this \emph{dual} score matching objective on the ImageNet64 dataset, and obtain a cross-entropy (negative log likelihood) value comparable to the state of the art. We further validate our approach by showing that our energy model \emph{strongly generalizes}: log probabilities estimated with two networks trained on non-overlapping data subsets are nearly identical. Finally, we demonstrate that both image probability and dimensionality of local neighborhoods vary substantially depending on image content, in contrast with conventional assumptions such as concentration of measure or support on a low-dimensional manifold.

Natural images often exhibit underlying sparse structures, with information density varying significantly across different spatial locations. However, most generative models rely on dense grid-based pixels or latents, neglecting this inherent sparsity. In this paper, we explore modeling visual generation paradigm via sparse non-grid latent representations. Specifically, we design a sparse autoencoder that represents an image as a small number of latents with their positional properties (i.e., regions of interest, RoIs) with high reconstruction quality. We then explore training flow-matching transformers jointly on non-grid latents and RoI values. To the best knowledge, we are the first to address spatial sparsity using RoIs in generative process. Experimental results show that our sparse flow-based transformers have competitive performance compared with dense grid-based counterparts with significantly reduced lower compute, and reaches a competitive 2.76 FID with just 64 latents on class-conditional ImageNet $256\times 256$ generation.

While text-to-image models like GPT-4o-Image and FLUX are rapidly proliferating, they often encounter challenges such as hallucination, bias, and the production of unsafe, low-quality output. To effectively address these issues, it is crucial to align these models with desired behaviors based on feedback from a multimodal judge. Despite their significance, current multimodal judges frequently undergo inadequate evaluation of their capabilities and limitations, potentially leading to misalignment and unsafe fine-tuning outcomes. To address this issue, we introduce MJ-Bench, a novel benchmark which incorporates a comprehensive preference dataset to evaluate multimodal judges in providing feedback for image generation models across six key perspectives: alignment, safety, image quality, bias, composition, and visualization. Specifically, we evaluate a large variety of multimodal judges including smaller-sized CLIP-based scoring models, open-source VLMs, and close-source VLMs on each decomposed subcategory of our preference dataset. Experiments reveal that close-source VLMs generally provide better feedback, with GPT-4o outperforming other judges in average. Compared with open-source VLMs, smaller-sized scoring models can provide better feedback regarding text-image alignment and image quality, while VLMs provide more accurate feedback regarding safety and generation bias due to their stronger reasoning capabilities. Further studies in feedback scale reveal that VLM judges can generally provide more accurate and stable feedback in natural language than numerical scales. Notably, human evaluations on end-to-end and fine-tuned models using separate feedback from these multimodal judges provide similar conclusions, further confirming the effectiveness of MJ-Bench.

Diffusion-based models have shown great promise in molecular generation but often require a large number of sampling steps to generate valid samples. In this paper, we introduce a novel Straight-Line Diffusion Model (SLDM) to tackle this problem, by formulating the diffusion process to follow a linear trajectory. The proposed process aligns well with the noise sensitivity characteristic of molecular structures and uniformly distributes reconstruction effort across the generative process, thus enhancing learning efficiency and efficacy. Consequently, SLDM achieves state-of-the-art performance on 3D molecule generation benchmarks, delivering a 100-fold improvement in sampling efficiency.

We propose a novel regularization loss that enforces standard Gaussianity, encouraging samples to align with a standard Gaussian distribution. This facilitates a range of downstream tasks involving optimization in the latent space of text-to-image models. We treat elements of a high-dimensional sample as one-dimensional standard Gaussian variables and define a composite loss that combines moment-based regularization in the spatial domain with power spectrum-based regularization in the spectral domain. Since the expected values of moments and power spectrum distributions are analytically known, the loss promotes conformity to these properties. To ensure permutation invariance, the losses are applied to randomly permuted inputs. Notably, existing Gaussianity-based regularizations fall within our unified framework: some correspond to moment losses of specific orders, while the previous covariance-matching loss is equivalent to our spectral loss but incurs higher time complexity due to its spatial-domain computation. We showcase the application of our regularization in generative modeling for test-time reward alignment with a text-to-image model, specifically to enhance aesthetics and text alignment. Our regularization outperforms previous Gaussianity regularization, effectively prevents reward hacking and accelerates convergence.

The scarcity of annotated medical imaging data has driven significant progress in semi-supervised learning to alleviate reliance on expensive expert labeling. While foundational vision models such as the Segment Anything Model (SAM) exhibit robust generalization in generic segmentation tasks, their direct application to medical images often results in suboptimal performance. To address this challenge, in this work, we propose a novel fully SAM-based semi-supervised medical image segmentation framework and develop the corresponding knowledge distillation-based learning strategy. Specifically, we first employ an efficient SAM variant as the backbone network of the semi‑supervised framework and update the default prompt embedding of SAM to unleash its full potential. Then, we utilize an original SAM, which is rich in prior knowledge, as the teacher to optimize our efficient student SAM backbone through hierarchical knowledge distillation and a dynamic loss weighting strategy. Extensive experiments on various medical datasets demonstrate that our method outperforms state-of-the-art semi-supervised segmentation approaches. Especially, our model requires less than 10% of the parameter size of the original SAM, enabling substantially lower deployment and storage overhead in real-world clinical settings.