ICLR 2026 Papers

Recent GRPO-based approaches built on flow matching models have shown remarkable improvements in human preference alignment for text-to-image generation. Nevertheless, they still suffer from the sparse reward problem: the terminal reward of the entire denoising trajectory is applied to all intermediate steps, resulting in a mismatch between the global feedback signals and the exact fine-grained contributions at intermediate denoising steps. To address this issue, we introduce \textbf{DenseGRPO}, a novel framework that aligns human preference with dense rewards, which evaluates the fine-grained contribution of each denoising step. Specifically, our approach includes two key components: (1) we propose to predict the step-wise reward gain as dense reward of each denoising step, which applies a reward model on the intermediate clean images via an ODE-based approach. This manner ensures an alignment between feedback signals and the contributions of individual steps, facilitating effective training; and (2) based on the estimated dense rewards, a mismatch drawback between the uniform exploration setting and the time-varying noise intensity in existing GRPO-based methods is revealed, leading to an inappropriate exploration space. Thus, we propose a reward-aware scheme to calibrate the exploration space by adaptively adjusting a timestep-specific stochasticity injection in the SDE sampler, ensuring a suitable exploration space at all timesteps. Extensive experiments on multiple standard benchmarks demonstrate the effectiveness of the proposed DenseGRPO and highlight the critical role of the valid dense rewards in flow matching model alignment.

Synthetic time series generation (TSG) is crucial for applications such as privacy preservation, data augmentation, and anomaly detection. A key challenge in TSG lies in modeling the multi-modal distributions of time series, which requires simultaneously capturing diverse high-level representation distributions and preserving local temporal fidelity. Most existing diffusion models, however, are constrained by their single-space focus: latent-space models capture representation distributions but often compromise local fidelity, while data-space models preserve local details in the data space but struggle to learn high-level representations essential for multi-modal time series. To address these limitations, we propose L2D-Diff, a dual-space diffusion framework for synthetic time series generation. Specifically, L2D-Diff first compresses input sequences into a latent space to efficiently model the distribution of time series representations. The distribution then guides a data-space diffusion model to refine local data details, enabling faithful generation of time series distribution without relying on external conditions. Experiments on both single-modal and multi-modal datasets demonstrate the effectiveness of L2D-Diff in tackling unconditional TSG tasks. Ablation studies further highlight the necessity and impact of its dual-space design, showcasing its capability to achieve representation coherence and local fidelity.

This paper investigates image inpainting with preference alignment. Instead of introducing a novel method, we go back to basics and revisit fundamental problems in achieving such alignment. We leverage the prominent direct preference optimization approach for alignment training and employ public reward models to construct preference training datasets. Experiments are conducted across nine reward models, two benchmarks, and two baseline models with varying structures and generative algorithms. Our key findings are as follows: (1) Most reward models deliver valid reward scores for constructing preference data, even if some of them are not reliable evaluators. (2) Preference data demonstrates robust trends in both candidate scaling and sample scaling across models and benchmarks. (3) Observable biases in reward models, particularly in brightness, composition, and color scheme, render them susceptible to cause reward hacking. (4) A simple ensemble of these models yields robust and generalizable results by mitigating such biases. Built upon these observations, our alignment models significantly outperform prior models across standard metrics, GPT-4 assessments, and human evaluations, without any changes to model structures or the use of new datasets. We hope our work can set a simple yet solid baseline, pushing this promising frontier. Our code is available at: https://github.com/shenytzzz/Follow-Your-Preference.

We introduce SANA-Video, a small diffusion model that can efficiently generate videos up to 720×1280 resolution and minute-length duration. SANA-Video synthesizes high-resolution, high-quality and long videos with strong text-video alignment at a remarkably fast speed, deployable on RTX 5090 GPU. Two core designs ensure our efficient, effective and long video generation: (1) Linear DiT: We leverage linear attention as the core operation, which is more efficient than vanilla attention given the large number of tokens processed in video generation. (2) Constant-Memory KV cache for Block Linear Attention: we design block-wise autoregressive approach for long video generation by employing a constant-memory state, derived from the cumulative properties of linear attention. This KV cache provides the Linear DiT with global context at a fixed memory cost, eliminating the need for a traditional KV cache and enabling efficient, minute-long video generation. In addition, we explore effective data filters and model training strategies, narrowing the training cost to 12 days on 64 H100 GPUs, which is only 1\% of the cost of MovieGen. Given its low cost, SANA-Video achieves competitive performance compared to modern state-of-the-art small diffusion models (e.g., Wan 2.1-1.3B and SkyReel-V2-1.3B) while being 16x faster in measured latency. Moreover, SANA-Video can be deployed on RTX 5090 GPUs with NVFP4 precision, accelerating the inference speed of generating a 5-second 720p video from 71s to 29s (2.4x} speedup). In summary, SANA-Video enables low-cost, high-quality video generation. Code and model will be publicly released.

Recent advances in video generation have enabled thrilling experiences in producing realistic videos driven by scalable diffusion transformers. However, they usually fail to produce satisfactory outputs that are aligned to users' authentic demands and preferences. In this work, we introduce Dual-Iterative Optimization (Dual-IPO), an iterative paradigm that sequentially optimizes both the reward model and the video generation model for improved synthesis quality and human preference alignment. For the reward model, our framework ensures reliable and robust reward signals via CoT-guided reasoning, voting-based self-consistency, and preference certainty estimation. Given this, we optimize video foundation models with guidance of signals from reward model's feedback, thus improving the synthesis quality in subject consistency, motion smoothness and aesthetic quality, etc. The reward model and video generation model complement each other and are progressively improved in the multi-round iteration, without requiring tediously manual preference annotations. Comprehensive experiments demonstrate that the proposed Dual-IPO can effectively and consistently improve the video generation quality of base model with various architectures and sizes, even help a model with only 2B parameters surpass a 5B one. Moreover, our analysis experiments and ablation studies identify the rational of our systematic design and the efficacy of each component.

Fast and high-quality language generation is the holy grail that people pursue in the age of AI. In this work, we introduce **Di**screte **Di**ffusion Divergence **Instruct** (**DiDi-Instruct**), a training-based method that initializes from a pre-trained diffusion large language model (dLLM) and distills a few-step student for fast generation. The model distilled with DiDi-Instruct matches or surpasses its dLLM teacher and the GPT-2 baseline while providing up to **64$\times$ acceleration**. The theoretical foundation of DiDi-Instruct is a novel framework based on integral KL-divergence minimization, which leads to a practical training algorithm. We further introduce *grouped reward normalization, intermediate-state matching, and the reward-guided ancestral sampler* to improve *training stability, model coverage, and inference quality*. On the OpenWebText benchmark, DiDi-Instruct achieves perplexity ranging from 62.2 (8 NFEs) to 18.4 (128 NFEs), outperforming prior accelerated dLLMs and the GPT-2 baseline. These gains incur a negligible entropy loss (around $1$%) and reduce additional training wall-clock time by **more than $20\times$** compared to competing dLLM distillation methods. We further validate the robustness and effectiveness of DiDi-Instruct through extensive ablation studies, model scaling, downstream task evaluations, and unconditional protein sequence generation. In conclusion, DiDi-Instruct enables efficient and effective distillation for language generation in the blink of an eye.

The widespread adoption of Low-Rank Adaptation (LoRA) for efficient fine-tuning of large language models has created demand for scalable parameter generation methods that can synthesize adaptation weights directly from task descriptions, avoiding costly task-specific training. We present LoRAGen, a structure-aware method for generating LoRA parameters from natural language descriptions. Through empirical analysis of LoRA libraries, we identify two key structural properties of LoRA parameter spaces: non-uniqueness of low-rank decomposition and heterogeneous weight distributions across network modules. These properties necessitate specialized parameter generation methods rather than general weight space learning approaches. LoRAGen employs a latent diffusion model with two innovations: weight-space supervision on full adaptation matrices to handle decomposition non-uniqueness, and a module-aware Mix-of-Experts decoder that adapts to module-specific weight distributions. Experiments show LoRAGen achieves 96.0\% performance relative to task-specific LoRAs on FLAN-T5-large and 72.7\% on Gemma-2-2B-Instruct for in-distribution tasks, while obtaining 40.2\% on zero-shot generation across unseen tasks—surpassing baselines by nearly 5\%. Our work establishes the first structure-aware approach to LoRA generation with insights into adaptation weight space geometry.

While thinking-aware generation aims to improve performance on complex tasks, we identify a critical failure mode where existing sequential, autoregressive approaches can paradoxically degrade performance due to error propagation. To systematically analyze this issue, we propose ParaBench, a new benchmark designed to evaluate both text and image output modalities. Our analysis using ParaBench reveals that this performance degradation is strongly correlated with poor alignment between the generated reasoning and the final image. To resolve this, we propose a parallel multimodal diffusion framework that enables continuous, bidirectional interaction between text and images throughout the entire denoising trajectory. The model is trained with supervised finetuning and then further optimized by Parallel Reinforcement Learning (ParaRL), a novel strategy that applies semantic rewards along the trajectory to enforce cross-modal consistency. Experiments validate that our approach significantly improves cross-modal alignment and semantic consistency, achieving a 6.9\% improvement in Output Alignment on ParaBench compared to the state-of-the-art model, Bagel, establishing a more robust paradigm for thinking-aware image synthesis.

Text-to-motion generation has advanced with diffusion models, yet existing systems often collapse complex multi-action prompts into a single embedding, leading to omissions, reordering, or unnatural transitions. In this work, we shift perspective by introducing a principled definition of an event as the smallest semantically self-contained action or state change in a text prompt that can be temporally aligned with a motion segment. Building on this definition, we pro- pose Event-T2M, a diffusion-based framework that decomposes prompts into events, encodes each with a motion-aware retrieval model, and integrates them through event-based cross-attention in Conformer blocks. Existing benchmarks mix simple and multi-event prompts, making it unclear whether models that succeed on single actions generalize to multi-action cases. To address this, we con- struct HumanML3D-E, the first benchmark stratified by event count. Experiments on HumanML3D, KIT-ML, and HumanML3D-E show that Event-T2M matches state-of-the-art baselines on standard tests while outperforming them as event complexity increases. Human studies validate the plausibility of our event definition, the reliability of HumanML3D-E, and the superiority of Event-T2M in generating multi-event motions that preserve order and naturalness close to ground- truth. These results establish event-level conditioning as a generalizable principle for advancing text-to-motion generation beyond single-action prompts.

We introduce EasyCreator, a novel 4D video creation framework capable of both generating and editing 4D content from a single monocular video input. By leveraging a powerful video inpainting foundation model as a generative prior, we reformulate 4D video creation as a video inpainting task, enabling the model to fill in missing content caused by camera trajectory changes or user edits. To facilitate this, we generate composite masked inpainting video data to effectively fine-tune the model for 4D video generation. Given an input video and its associated camera trajectory, we first perform depth-based point cloud rendering to obtain invisibility masks that indicate the regions that should be completed. Simultaneously, editing masks are introduced to specify user-defined modifications, and these are combined with the invisibility masks to create a composite masks dataset. During training, we randomly sample different types of masks to construct diverse and challenging inpainting scenarios, enhancing the model’s generalization and robustness in various 4D editing and generation tasks. To handle temporal consistency under large camera motion, we design a self-iterative tuning strategy that gradually increases the viewing angles during training, where the model is used to generate the next-stage training data after each fine-tuning iteration. Moreover, we introduce a temporal packaging module during inference to enhance generation quality. Our method effectively leverages the prior knowledge of the base model without degrading its original performance, enabling the generation of 4D videos with consistent multi-view coherence. In addition, our approach supports prompt-based content editing, demonstrating strong flexibility and significantly outperforming state-of-the-art methods in both quality and versatility.

Recent diffusion-based Multimodal Large Language Models (dMLLMs) suffer from high inference latency and therefore rely on caching techniques to accelerate decoding. However, the application of cache mechanisms often introduces undesirable repetitive text generation, a phenomenon we term the Repeat Curse. To better investigate underlying mechanism behind this issue, we analyze repetition generation through the lens of information flow. Our work reveals three key findings: (1) context tokens aggregate semantic information as anchors and guide the final predictions; (2) as information propagates across layers, the entropy of context tokens converges in deeper layers, reflecting the model’s growing prediction certainty; (3) Repetition is typically linked to disruptions in the information flow of context tokens and to the inability of their entropy to converge in deeper layers. Based on these insights, we present CoTA, a plug-and-play method for mitigating repetition. CoTA enhances the attention of context tokens to preserve intrinsic information flow patterns, while introducing a penalty term to the confidence score during decoding to avoid outputs driven by uncertain context tokens. With extensive experiments, CoTA demonstrates significant effectiveness in alleviating repetition and achieves consistent performance improvements on general tasks. Code will be made available.

Classifier-Free Guidance (CFG) is a widely used technique for conditional generation and improving sample quality in continuous diffusion models, and its extensions to discrete diffusion has recently started to be investigated. In order to improve the algorithms in a principled way, this paper starts by analyzing the exact effect of CFG in the context of a low-dimensional masked diffusion model, with a special emphasis on the guidance schedule. Our analysis shows that high guidance early in sampling (when inputs are heavily masked) harms generation quality, while late-stage guidance has a larger effect. These findings provide a theoretical explanation for empirical observations in recent studies on guidance schedules. The analysis also reveals an imperfection of the current CFG implementations. These implementations can unintentionally cause imbalanced transitions, such as unmasking too rapidly during the early stages of generation, which degrades the quality of the resulting samples. To address this, we draw insight from the analysis and propose a novel classifier-free guidance mechanism. Intuitively, our method smoothens the transport between the data distribution and the initial (masked) distribution, which results in improved sample quality. Remarkably, our method is achievable via a simple one-line code change. Experiments on conditional image and text generation empirically confirm the efficacy of our method.

Information-theoretic measures, such as Mutual Information (MI), play a crucial role in understanding non-linear relationships between random variables and are widely used across scientific disciplines. Yet, their use on real-world discrete data remains challenging. Existing methods typically rely on embedding discrete data into a continuous space and apply neural estimators originally designed for continuous distributions. This process requires careful engineering for both the embedding model and estimator architecture, but suffers from issues related to high data dimensionality. In this work, we introduce InfoSEDD, a discrete diffusion–based approach that bridges information-theoretic estimation and generative modeling such that they can be used to compute Kullback–Leibler divergences. Backed by Continuous Time Markov Chains theory principles, the design of InfoSEDD is lightweight and scalable and allows seamless integration with pretrained models. We showcase the versatility of our approach through applications on motif discovery in genetic promoter data, semantic-aware model selection in text summarization, and entropy estimation in Ising models. Finally, we construct consistency tests on real-world textual and genomics data. Our experiments demonstrate that InfoSEDD outperforms alternatives that rely on the ''embedding trick''. Our results position InfoSEDD as a robust and scalable tool for information-theoretic analysis of discrete data.

A hallmark of modern large-scale machine learning techniques is the use of training objectives that provide dense supervision to intermediate computations, such as teacher forcing the next token in language models or denoising step-by-step in diffusion models. This enables models to learn complex functions in a generalizable manner. Motivated by this observation, we investigate the benefits of iterative computation for temporal difference (TD) methods in reinforcement learning (RL). Typically, they represent value functions in a monolithic fashion, without iterative compute. We introduce floq (flow-matching Q-functions), an approach that parameterizes the Q-function using a velocity field and trains it with techniques from flow-matching, typically used in generative modeling. This velocity field underneath the flow is trained using a TD-learning objective, which bootstraps from values produced by a target velocity field, computed by running multiple steps of numerical integration. Crucially, floq allows for more fine-grained control and scaling of the Q-function capacity than monolithic architectures, by appropriately setting the number of integration steps. Across a suite of challenging offline RL benchmarks and online fine-tuning tasks, floq improves performance by nearly 1.8x. floq scales capacity far better than standard TD-learning architectures, highlighting the potential of iterative computation for value learning.

Recent advances in diffusion-based image generation models (IGMs), such as Stable Diffusion (SD), have substantially improved the quality and diversity of AI-generated content. However, these models also pose ethical, legal, and societal risks, including the generation of harmful, misleading, or copyright-infringing material. Machine unlearning (MU) has emerged as a promising mitigation by selectively removing undesirable concepts from pretrained models, yet the robustness of existing methods, particularly under multi-modal adversarial inputs, remains insufficiently explored. To address this gap, we propose RECALL, a multi-modal adversarial framework for systematically evaluating and compromising the robustness of unlearned IGMs. Unlike prior approaches that primarily optimize adversarial text prompts, RECALL exploits the native multi-modal conditioning of diffusion models by efficiently optimizing adversarial image prompts guided by a single semantically relevant reference image. Extensive experiments across ten state-of-the-art unlearning methods and diverse representative tasks show that RECALL consistently surpasses existing baselines in adversarial effectiveness, computational efficiency, and semantic fidelity to the original prompt. These results reveal critical vulnerabilities in current unlearning pipelines and underscore the need for more robust, verifiable unlearning mechanisms. More than just an attack, RECALL also serves as an auditing tool for model owners and unlearning practitioners, enabling systematic robustness evaluation. Code and data are available at https://anonymous.4open.science/r/RECALL.

Speech tokenizers are foundational to speech language models, yet existing approaches face two major challenges: (1) balancing trade-offs between encoding semantics for understanding and acoustics for reconstruction, and (2) achieving low bit rates and low token rates. We propose Speech Diffusion Tokenizer (SiTok), a diffusion autoencoder that jointly learns semantic-rich representations through supervised learning and enables high-fidelity audio reconstruction with diffusion. We scale SiTok to 1.6B parameters and train it on 2 million hours of speech. Experiments show that SiTok outperforms strong baselines on both reconstruction and understanding tasks, at an extremely low token rate of 12.5 Hz and a bit-rate of 200 bits-per-second.

Adversarial Training (AT) is a known, powerful, well-established technique for improving classifier robustness to input perturbations, yet its applicability beyond discriminative settings remains limited. Motivated by the widespread use of score-based generative models and their need to operate robustly under substantial noisy or corrupted input data, we propose an adaptation of AT for these models, providing a thorough empirical assessment. We introduce a principled formulation of AT for Diffusion Models (DMs) that replaces the conventional *invariance* objective with an *equivariance* constraint aligned to the denoising dynamics of score matching. Our method integrates seamlessly into diffusion training by adding either random perturbations--similar to randomized smoothing--or adversarial ones--akin to AT. Our approach offers several advantages: **(a)** tolerance to heavy noise and corruption, **(b)** reduced memorization, **(c)** robustness to outliers and extreme data variability and **(d)** resilience to iterative adversarial attacks. We validate these claims on proof-of-concept low- and high-dimensional datasets with *known* ground-truth distributions, enabling precise error analysis. We further evaluate on standard benchmarks (CIFAR-10, CelebA, and LSUN Bedroom), where our approach shows improved robustness and preserved sample fidelity under severe noise, data corruption, and adversarial evaluation. Code available upon acceptance.

Diffusion generative models have demonstrated remarkable empirical success in the recent years and are now considered the state-of-the-art generative models in modern AI. These models consist of a forward process, which gradually diffuses the data distribution to a noise distribution spanning the whole space, and a backward process, which inverts this transformation to recover the data distribution from noise. Most of the existing literature assumes that the underlying space is Euclidean. However, in many practical applications, the data are constrained to lie on a submanifold of Euclidean space. Addressing this setting, de Bortoli et al. (2022) introduced Riemannian diffusion models and proved that using an exponentially small step size yields small sampling error in Wasserstein distance, provided the data distribution is smooth and strictly positive. In this paper, we prove that a *polynomially small stepsize* suffices to guarantee small sampling error in total variation distance, without any assumption on the smoothness or positivity of the data distribution. Our analysis only requires mild and standard curvature assumptions on the underlying manifold. The main ingredients in our proof are Li-Yau estimate for log-gradient of heat kernel, and Minakshisundaram-Pleijel parametrix expansion for perturbed heat equation. Our approach opens the door to a sharper analysis of diffusion models on non-Euclidean spaces.

In-context learning lets large models adapt to new tasks from a few demonstrations, but it has shown limited success in molecular design, where labeled data are scarce and properties span millions of biological assays and material measurements. We introduce demonstration-conditioned diffusion models (DemoDiff), which define task contexts through molecule–score examples instead of texts. These demonstrations guide a denoising Transformer to generate molecules aligned with target properties. For scalable pretraining, we develop a new molecular tokenizer with Node Pair Encoding that represents molecules at the motif level, requiring 5.5$\times$ fewer nodes. We pretrain a 0.7B parameter model on datasets covering drugs and materials. Across 33 design tasks in six categories, DemoDiff matches or surpasses language models 100–1000$\times$ larger and achieves an average rank of 4.10 compared to 6.56–17.95 for 19 baselines. These results position DemoDiff as a molecular foundation model for in-context molecular design.