ICLR 2026 Papers

Classifier-Free Guidance (CFG) is widely used in diffusion and flow-based generative models for high-quality conditional generation, yet its theoretical properties remain incompletely understood. By connecting CFG to the high-dimensional framework of diffusion regimes, we show that in sufficiently high dimensions it reproduces the correct target distribution—a “blessing-of-dimensionality” result. Leveraging this theoretical framework, we analyze how the well-known artifacts of mean overshoot and variance shrinkage emerge in lower dimensions, characterizing how they become more pronounced as dimensionality decreases. Building on these insights, we propose a simple nonlinear extension of CFG, proving that it mitigates both effects while preserving CFG’s practical benefits. Finally, we validate our approach through numerical simulations on Gaussian mixtures and real-world experiments on diffusion and flow-matching state-of-the-art class-conditional and text-to-image models, demonstrating continuous improvements in sample quality, diversity, and consistency.

Diffusion models have shown impressive performance in many visual generation and manipulation tasks. Many existing methods focus on training a model for a specific task, especially, text-to-video (T2V) generation, while many other works focus on finetuning the pretrained T2V model for image-to-video (I2V), video-to-video (V2V), image and video manipulation tasks, \etc. However, training a strong T2V foundation model requires a large amount of high-quality annotations, which is very costly. In addition, many existing models can perform only one or several tasks. In this work, we introduce a unified framework, namely \textit{many-for-many}, which leverages the available training data from many different visual generation and manipulation tasks to train a single model for those different tasks. Specifically, we design a lightweight adapter to unify the different conditions in different tasks, then employ a joint image-video learning strategy to progressively train the model from scratch. Our joint learning not only leads to a unified generation and manipulation model but also benefits the performance of different tasks. In addition, we introduce depth maps as a condition to help our model better perceive the 3D space in visual generation. Two versions of our model are trained with different model sizes (8B and 2B), each of which can perform more than 10 different tasks. In particular, our 8B model demonstrates highly competitive performance in different generation and manipulation tasks compared to open-source and even commercial engines. Our models and source codes will be made publicly available.

Diffusion-based generative models have emerged as highly effective methods for synthesizing high-quality samples. Recent works have focused on analyzing the convergence of their generation process with minimal assumptions, either through reverse SDEs or probability flow ODEs. The best known guarantees, without any smoothness assumptions, for the KL divergence so far achieve a linear dependence on the data dimension $d$ and an inverse quadratic dependence on accuracy level $\varepsilon$. In this work, we present a refined analysis for the standard Exponential Integrator discretization that improves the dependence on $\varepsilon$, at the same time maintaining the linear dependence on $d$. Following recent works on higher order/randomized midpoint discretizations, we model the generation process as a composition of two steps: a reverse ODE step followed by a smaller noising step, which leads to better dependence on step size. We then provide a novel analysis which achieves linear dependence on $d$ for the ODE discretization error without any smoothness assumptions. Specifically, we introduce a general ODE-based counterpart of the stochastic localization argument from Benton et al and develop new proof techniques to bound second-order spatial derivatives of the score function -- terms that do not arise in previous diffusion analyses and cannot be handled by existing techniques. Leveraging this framework, we prove that $\tilde{O}\left(\tfrac{d \log^{3/2}(1/\delta)}{\varepsilon}\right)$ steps suffice to approximate the target distribution—corrupted by Gaussian noise of variance $\delta$—to within $O(\varepsilon^2)$ in KL divergence, improving upon the previous best result requiring $\tilde{O}\left(\tfrac{d \log^2(1/\delta)}{\varepsilon^2}\right)$ steps.

Diffusion Policy has demonstrated strong visuomotor modeling capabilities, but its high computational cost renders it impractical for real-time robotic control. Despite huge redundancy across repetitive denoising steps, existing diffusion acceleration techniques fail to generalize to Diffusion Policy due to fundamental architectural and data divergences. In this paper, we propose **B**lock-wise **A**daptive **C**aching (**BAC**), a method to accelerate Diffusion Policy by caching intermediate action features. BAC achieves lossless action generation acceleration by adaptively updating and reusing cached features at the block level, based on a key observation that feature similarities exhibit non-uniform temporal dynamics and distinct block-specific patterns. To operationalize this insight, we first design an Adaptive Caching Scheduler to identify optimal update timesteps by maximizing the global feature similarities between cached and skipped features. However, applying this scheduler for each block leads to significant error surges due to the inter-block propagation of caching errors, particularly within Feed-Forward Network (FFN) blocks. To mitigate this issue, we develop the Bubbling Union Algorithm, which truncates these errors by updating the upstream blocks with significant caching errors before downstream FFNs. As a training-free plugin, BAC is readily integrable with existing transformer-based Diffusion Policy and vision-language-action models. Extensive experiments on multiple robotic benchmarks demonstrate that BAC achieves up to 3x inference speedup for free. Project page: https://block-wise-adaptive-caching.github.io.

Recent Text-to-Video (T2V) models have demonstrated powerful capability in visual simulation of real-world geometry and physical laws, indicating its potential as implicit world models. Inspired by this, we explore the feasibility of leveraging the video generation prior for viewpoint planning from given 4D scenes, since videos internally accompany dynamic scenes with natural viewpoints. To this end, we propose a two-stage paradigm to adapt pre-trained T2V models for viewpoint prediction, in a compatible manner. First, we inject the 4D scene representation into the pre-trained T2V model via an adaptive learning branch, where the 4D scene is viewpoint-agnostic and the conditional generated video embeds the viewpoints visually. Then, we formulate viewpoint extraction as a hybrid-condition guided camera extrinsic denoising process. Specifically, a camera extrinsic diffusion branch is further introduced onto the pre-trained T2V model, by taking the generated video and 4D scene as input. Experimental results show the superiority of our proposed method over existing competitors, and ablation studies validate the effectiveness of our key technical designs. To some extent, this work proves the potential of video generation models toward 4D interaction in real world.

The Distribution Matching Distillation (DMD) has been successfully applied to text-to-image diffusion models such as Stable Diffusion (SD) 1.5. However, vanilla DMD suffers from convergence difficulties on large-scale flow-based text-to-image models, such as SD 3.5 and FLUX. In this paper, we first analyze the issues when applying vanilla DMD on large-scale models. Then, to overcome the scalability challenge, we propose implicit distribution alignment (IDA) to constrain the divergence between the generator and the fake distribution. Furthermore, we propose intra-segment guidance (ISG) to relocate the timestep denoising importance from the teacher model. With IDA alone, DMD converges for SD 3.5; employing both IDA and ISG, DMD converges for SD 3.5 and FLUX.1 dev. Together with a scaled VFM-based discriminator, our final model, dubbed **SenseFlow**, achieves superior performance in distillation for both diffusion based text-to-image models such as SDXL, and flow-matching models such as SD 3.5 Large and FLUX.1 dev. The source code will be released.

Various deep generative models have been proposed to estimate potential outcomes distributions from observational data. However, none of them have the favorable theoretical property of general Neyman-orthogonality and, associated with it, quasi-oracle efficiency and double robustness. In this paper, we introduce a general suite of generative Neyman-orthogonal (doubly-robust) learners that estimate the conditional distributions of potential outcomes. Our proposed generative doubly-robust learners (GDR-learners) are flexible and can be instantiated with many state-of-the-art deep generative models. In particular, we develop GDR-learners based on (a) conditional normalizing flows (which we call GDR-CNFs), (b) conditional generative adversarial networks (GDR-CGANs), (c) conditional variational autoencoders (GDR-CVAEs), and (d) conditional diffusion models (GDR-CDMs). Unlike the existing methods, our GDR-learners possess the properties of quasi-oracle efficiency and rate double robustness, and are thus asymptotically optimal. In a series of (semi-)synthetic experiments, we demonstrate that our GDR-learners are very effective and outperform the existing methods in estimating the conditional distributions of potential outcomes.

Large language models (LLMs) deliver remarkable performance but are costly to deploy, motivating knowledge distillation (KD) for efficient inference. Existing KD objectives typically match student and teacher probabilities via softmax, which blurs valuable logit information. While direct logit distillation (DLD) mitigates softmax smoothing, it fails to account for logit shift invariance, thereby restricting the solution space. We propose Concrete Score Distillation (CSD), a discrete score-matching objective that overcomes both softmax-induced smoothing and restrictions on the optimal solution set. We resolve the training instability and quadratic complexity of discrete score-matching in autoregressive LLMs, and the resulting CSD objective aligns relative logit differences across all vocabulary pairs between student and teacher with flexible weighting. We provide both mode-seeking and mode-covering instances within our framework and evaluate CSD on task-agnostic instruction-following, task-specific, and general chat capability distillation using GPT-2-1.5B, OpenLLaMA-7B, and Gemma-7B-IT, Qwen2.5-7B-IT, and Gemma2-9B-IT teachers. Experiments show that CSD consistently surpasses recent KD objectives, achieves favorable fidelity–diversity trade-offs, and yields complementary gains when combined with on-policy techniques, demonstrating its scalability and effectiveness for LLM distillation.

Generating high-resolution 3D CT volumes with fine details remains challenging due to substantial computational demands and optimization difficulties inherent to existing generative models. In this paper, we propose the Pixel-Level Residual Diffusion Transformer (PRDiT), a scalable generative framework that synthesizes high-quality 3D medical volumes directly at voxel-level. PRDiT introduces a two-stage training architecture comprising 1) a local denoiser in the form of an MLP-based blind estimator operating on overlapping 3D patches to separate low-frequency structures efficiently, and 2) a global residual diffusion transformer employing memory-efficient attention to model and refine high-frequency residuals across entire volumes. This coarse-to-fine modeling strategy simplifies optimization, enhances training stability, and effectively preserves subtle structures without the limitations of an autoencoder bottleneck. Extensive experiments conducted on the LIDC-IDRI and RAD-ChestCT datasets demonstrate that PRDiT consistently outperforms state-of-the-art models, such as HA-GAN, 3D LDM and WDM-3D, achieving significantly lower 3D FID, MMD and Wasserstein distance scores.

In-context image editing aims to modify images based on a contextual sequence comprising text and previously generated images. Existing methods typically depend on task-specific pipelines and expert models (e.g., segmentation and inpainting) to curate training data. In this work, we explore whether an in-context image editing model can be learned directly from videos. We introduce a scalable approach to annotate videos as interleaved multimodal sequences. To effectively learn from this data, we design three proxy tasks: next-image prediction, current segmentation prediction, and next-segmentation prediction. Additionally, we propose a novel multi-turn image editing benchmark to advance research in this area. Extensive experiments demonstrate that our model exhibits strong in-context image editing capabilities and achieves state-of-the-art results on two multi-turn image editing benchmarks. Despite being trained exclusively on videos, our model also shows promising abilities in multi-concept composition, story generation, and chain-of-editing applications.

Autoregressive decoding in large language models (LLMs) requires $\mathcal{O}(n)$ sequential steps for $n$ tokens, fundamentally limiting inference throughput. Recent diffusion-based LLMs (dLLMs) enable parallel token generation through iterative denoising. However, current parallel decoding strategies rely on fixed, input-agnostic heuristics (e.g., confidence thresholds), which fail to adapt to input-specific characteristics, resulting in suboptimal speed-quality trade-offs across diverse NLP tasks. In this work, we explore a more flexible and dynamic approach to parallel decoding. We propose **Learning to Parallel Decode (Learn2PD)**, a framework that trains a lightweight and adaptive filter model to predict, for each token position, whether the current prediction matches the final output. This learned filter approximates an oracle parallel decoding strategy that unmasks tokens only when correctly predicted. Importantly, the filter model is learned in a post-training manner, requiring only a small amount of computation to optimize it (minute-level GPU time). Additionally, we introduce **End-of-Text Prediction (EoTP)** to detect decoding completion at the end of sequence, avoiding redundant decoding of padding tokens. Experiments on the LLaDA benchmark demonstrate that our method achieves up to **22.58×** speedup without any performance drop, and up to **57.51×** when combined with KV-Cache.

Generative models have emerged as powerful tools for planning, with compositional approaches offering particular promise for modeling long-horizon task distributions by composing together local, modular generative models. This compositional paradigm spans diverse domains, from multi-step manipulation planning to panoramic image synthesis to long video generation. However, compositional generative models face a critical challenge: when local distributions are multimodal, existing composition methods average incompatible modes, producing plans that are neither locally feasible nor globally coherent. We propose Compositional Diffusion with Guided Search (CDGS), which addresses this \emph{mode averaging} problem by embedding search directly within the diffusion denoising process. Our method explores diverse combinations of local modes through population-based sampling, prunes infeasible candidates using likelihood-based filtering, and enforces global consistency through iterative resampling between overlapping segments. CDGS matches oracle performance on seven robot manipulation tasks, outperforming baselines that lack compositionality or require long-horizon training data. The approach generalizes across domains, enabling coherent text-guided panoramic images and long videos through effective local-to-global message passing. More details: https://cdgsearch.github.io/

Classical search algorithms have long underpinned modern artificial intelligence. In this work, we tackle the challenge of inference-time control in diffusion models—adapting generated outputs to meet diverse test-time objectives—using principles from classical search. We propose a general framework that orchestrates local and global search to efficiently navigate the generative space. It performs compute-efficient global exploration using breadth-first and depth-first tree search and employs a theoretically grounded, scalable local search via annealed Langevin MCMC. We evaluate our approach on a range of challenging domains, including planning, offline reinforcement learning, and image generation, and observe significant gains in both performance and efficiency over baseline methods. These results demonstrate that classical search offers a principled and practical foundation for inference-time scaling in diffusion models. By jointly scaling local and global search for the first time, our framework establishes a new Pareto frontier across challenging decision-making domains.

Fast, unconditional 3D generative models can now produce high-quality molecules, but adapting them for specific design tasks often requires costly retraining. To address this, we introduce Interpolate-Integrate and Replacement Guidance, two training-free, inference-time conditioning strategies that provide control over E(3)-equivariant flow-matching models. Our methods generate bioisosteric 3D molecules by conditioning on seed ligands or fragment sets to preserve key determinants like shape and pharmacophore patterns, without requiring the original fragment atoms to be present. We demonstrate their effectiveness on three drug-relevant tasks: natural product ligand hopping, bioisosteric fragment merging, and pharmacophore merging.

We introduce $\texttt{PairFlow}$, a lightweight preprocessing step for training Discrete Flow Models (DFMs) to achieve few-step sampling without requiring a pretrained teacher. DFMs have recently emerged as a new class of generative models for discrete data, offering strong performance. However, they suffer from slow sampling due to their iterative nature. Existing acceleration methods largely depend on finetuning, which introduces substantial additional training overhead. $\texttt{PairFlow}$ addresses this issue with a lightweight preprocessing step. Inspired by ReFlow and its extension to DFMs, we train DFMs from coupled samples of source and target distributions, without requiring any pretrained teacher. At the core of our approach is a closed-form inversion for DFMs, which allows efficient construction of paired source–target samples. Despite its extremely low cost, taking only up to 1.7\% of the compute needed for full model training, $\texttt{PairFlow}$ matches or even surpasses the performance of two-stage training involving finetuning. Furthermore, models trained with our framework provide stronger base models for subsequent distillation, yielding further acceleration after finetuning. Experiments on molecular data as well as binary and RGB images demonstrate the broad applicability and effectiveness of our approach.

Erasing concepts from large-scale text-to-image (T2I) diffusion models has become increasingly crucial due to the growing concerns over copyright infringement, offensive content, and privacy violations. In scalable applications, fine-tuning-based methods are time-consuming to precisely erase multiple target concepts, while real-time editing-based methods often degrade the generation quality of non-target concepts due to conflicting optimization objectives. To address this dilemma, we introduce SPEED, an efficient concept erasure approach that directly edits model parameters. SPEED searches for a null space, a model editing space where parameter updates do not affect non-target concepts, to achieve scalable and precise erasure. To facilitate accurate null space optimization, we incorporate three complementary strategies: Influence-based Prior Filtering (IPF) to selectively retain the most affected non-target concepts, Directed Prior Augmentation (DPA) to enrich the filtered retain set with semantically consistent variations, and Invariant Equality Constraints (IEC) to preserve key invariants during the T2I generation process. Extensive evaluations across multiple concept erasure tasks demonstrate that SPEED consistently outperforms existing methods in non-target preservation while achieving efficient and high-fidelity concept erasure, successfully erasing 100 concepts within only 5 seconds.

In this work, we propose aligning pretrained visual encoders to serve as tokenizers for latent diffusion models in image generation. Unlike training a variational autoencoder (VAE) from scratch, which primarily emphasizes low-level details, our approach leverages the rich semantic structure of foundation encoders. We introduce a three-stage alignment strategy called AlignTok: (1) freeze the encoder and train an adapter and a decoder to establish a semantic latent space; (2) jointly optimize all components with an additional semantic preservation loss, enabling the encoder to capture perceptual details while retaining high-level semantics; and (3) refine the decoder for improved reconstruction quality. This alignment yields semantically rich image tokenizers that benefit diffusion models. On ImageNet 256$\times$256, our tokenizer accelerates the convergence of diffusion models, reaching a gFID of 1.90 within just 64 epochs, and improves generation both with and without classifier-free guidance. Scaling to LAION, text-to-image models trained with our tokenizer consistently outperforms FLUX VAE and VA-VAE under the same training steps. Overall, our method is simple, scalable, and establishes a semantically grounded paradigm for continuous tokenizer design.

MeanFlow has recently emerged as a powerful framework for few-step generative modeling trained from scratch, but its success is not yet fully understood. In this work, we show that the MeanFlow objective naturally decomposes into two parts: trajectory flow matching and trajectory consistency. Through gradient analysis, we find that these terms are strongly negatively correlated, causing optimization conflict and slow convergence. Motivated by these insights, we introduce $\alpha$-Flow, a broad family of objectives that unifies trajectory flow matching, Shortcut Model, and MeanFlow under one formulation. By adopting a curriculum strategy that smoothly anneals from trajectory flow matching to MeanFlow, $\alpha$-Flow disentangles the conflicting objectives, and achieves better convergence. When trained from scratch on class-conditional ImageNet-1K 256×256 with vanilla DiT backbones, $\alpha$-Flow consistently outperforms MeanFlow across scales and settings. Our largest $\alpha$-Flow-XL/2+ model achieves new state-of-the-art results using vanilla DiT backbones, with FID scores of 2.58 (1-NFE) and 2.15 (2-NFE). The source code and pre-trained checkpoints will be publicly released.

Diffusion Transformers (DiTs) deliver state-of-the-art generative performance but their quadratic training cost with sequence length makes large-scale pretraining prohibitively expensive. Token dropping can reduce training cost, yet naïve strategies degrade representations, and existing methods are either parameter-heavy or fail at high drop ratios. We present SPRINT (Sparse--Dense Residual Fusion for Efficient Diffusion Transformers), a simple method that enables aggressive token dropping (up to 75%) while preserving quality. SPRINT leverages the complementary roles of shallow and deep layers: early layers process all tokens to capture local detail, deeper layers operate on a sparse subset to cut computation, and their outputs are fused through residual connections. Training follows a two-stage schedule: long masked pre-training for efficiency followed by short full-token fine-tuning to close the train--inference gap. On ImageNet-1K 256^2, SPRINT achieves 9.8x training savings with comparable FID/FDD, and at inference, its Path-Drop Guidance (PDG) nearly halves FLOPs while improving quality. These results establish SPRINT as a simple, effective, and general solution for efficient DiT training.