ICML 2024 Papers

Minimum Bayes Risk (MBR) decoding has been shown to be a powerful alternative to beam search decoding in a variety of text generation tasks. MBR decoding selects a hypothesis from a pool of hypotheses that has the least expected risk under a probability model according to a given utility function. Since it is impractical to compute the expected risk exactly over all possible hypotheses, two approximations are commonly used in MBR. First, it integrates over a sampled set of hypotheses rather than over all possible hypotheses. Second, it estimates the probability of each hypothesis using a Monte Carlo estimator. While the first approximation is necessary to make it computationally feasible, the second is not essential since we typically have access to the model probability at inference time. We propose model-based MBR (MBMBR), a variant of MBR that uses the model probability itself as the estimate of the probability distribution instead of the Monte Carlo estimate. We show analytically and empirically that the model-based estimate is more promising than the Monte Carlo estimate in text generation tasks. Our experiments show that MBMBR outperforms MBR in several text generation tasks, both with encoder-decoder models and with language models.

Data plays a pivotal role in the development of both classic and learning-based methods for Mixed-Integer Linear Programming (MILP). However, the scarcity of data in real-world applications underscores the necessity for MILP instance generation methods. Currently, these methods primarily rely on iterating random single-constraint modifications, disregarding the underlying problem structure with constraint interrelations, thereby leading to compromised quality and solvability. In this paper, we propose ACM-MILP, a framework for MILP instance generation, to achieve adaptive constraint modification and constraint interrelation modeling. It employs an adaptive constraint selection mechanism based on probability estimation within the latent space to preserve instance characteristics. Meanwhile, it detects and groups strongly related constraints through community detection, enabling collective modifications that account for constraint dependencies. Experimental results show significant improvements in problem-solving hardness similarity under our framework. Additionally, in the downstream task, we showcase the efficacy of our generated instances for hyperparameter tuning. Source code is available: https://github.com/Thinklab-SJTU/ACM-MILP.

Contrastive Language-Image Pretraining (CLIP) has achieved remarkable success, leading to rapid advancements in multimodal studies. However, CLIP faces a notable challenge in terms of *inefficient data utilization*. It relies on a single contrastive supervision for each image-text pair during representation learning, disregarding a substantial amount of valuable information that could offer richer supervision. Additionally, the retention of non-informative tokens leads to increased computational demands and time costs, particularly in CLIP's ViT image encoder. To address these issues, we propose **M**ulti-Perspective **L**anguage-**I**mage **P**retraining (**MLIP**). In MLIP, we leverage the frequency transform's sensitivity to both high and low-frequency variations, which complements the spatial domain's sensitivity limited to low-frequency variations only. By incorporating frequency transforms and token-level alignment, we expand CILP's single supervision into multi-domain and multi-level supervision, enabling a more thorough exploration of informative image features. Additionally, we introduce a token merging method guided by comprehensive semantics from the frequency and spatial domains. This allows us to merge tokens to multi-granularity tokens with a controllable compression rate to accelerate CLIP. Extensive experiments validate the effectiveness of our design.

Diffusion models (DMs) have revolutionized generative learning. They utilize a diffusion process to encode data into a simple Gaussian distribution. However, encoding a complex, potentially multimodal data distribution into a single *continuous* Gaussian distribution arguably represents an unnecessarily challenging learning problem. We propose ***Dis**crete-**Co**ntinuous Latent Variable **Diff**usion Models (DisCo-Diff)* to simplify this task by introducing complementary *discrete* latent variables. We augment DMs with learnable discrete latents, inferred with an encoder, and train DM and encoder end-to-end. DisCo-Diff does not rely on pre-trained networks, making the framework universally applicable. The discrete latents significantly simplify learning the DM's complex noise-to-data mapping by reducing the curvature of the DM's generative ODE. An additional autoregressive transformer models the distribution of the discrete latents, a simple step because DisCo-Diff requires only few discrete variables with small codebooks. We validate DisCo-Diff on toy data, several image synthesis tasks as well as molecular docking, and find that introducing discrete latents consistently improves model performance. For example, DisCo-Diff achieves state-of-the-art FID scores on class-conditioned ImageNet-64/128 datasets with ODE sampler.

To ensure that text generated by large language models (LLMs) is in an expected format, constrained decoding methods propose to enforce strict formal language constraints during generation. However, as we show in this work, not only do such methods often incur performance overhead during generation, but many of them also significantly impair task accuracy, if they do not correctly align the underlying LLM sub-word vocabularies with external constraints. To address this, we present a novel decoding algorithm, DOMINO, that can enforce constraints in a fully subword-aligned fashion, while leveraging pre-computation and speculative decoding to achieve virtually no overhead and in some cases even almost 2$\times$ speedup over unconstrained decoding -- thereby outperforming existing approaches by a wide margin. We release DOMINO as open source at https://github.com/eth-sri/domino.

Adversarial camouflage is a widely used physical attack against vehicle detectors for its superiority in multi-view attack performance. One promising approach involves using differentiable neural renderers to facilitate adversarial camouflage optimization through gradient back-propagation. However, existing methods often struggle to capture environmental characteristics during the rendering process or produce adversarial textures that can precisely map to the target vehicle, resulting in suboptimal attack performance. Moreover, these approaches neglect diverse weather conditions, reducing the efficacy of generated camouflage across varying weather scenarios. To tackle these challenges, we propose a robust and accurate camouflage generation method, namely RAUCA. The core of RAUCA is a novel neural rendering component, Neural Renderer Plus (NRP), which can accurately project vehicle textures and render images with environmental characteristics such as lighting and weather. In addition, we integrate a multi-weather dataset for camouflage generation, leveraging the NRP to enhance the attack robustness. Experimental results on six popular object detectors show that RAUCA consistently outperforms existing methods in both simulation and real-world settings.

Diffusion models exhibit powerful generative capabilities enabling noise mapping to data via reverse stochastic differential equations. However, in image restoration, the focus is on the mapping relationship from low-quality to high-quality images. Regarding this issue, we introduce the Generalized Ornstein-Uhlenbeck Bridge (GOUB) model. By leveraging the natural mean-reverting property of the generalized OU process and further eliminating the variance of its steady-state distribution through the Doob's *h*–transform, we achieve diffusion mappings from point to point enabling the recovery of high-quality images from low-quality ones. Moreover, we unravel the fundamental mathematical essence shared by various bridge models, all of which are special instances of GOUB and empirically demonstrate the optimality of our proposed models. Additionally, we present the corresponding Mean-ODE model adept at capturing both pixel-level details and structural perceptions. Experimental outcomes showcase the state-of-the-art performance achieved by both models across diverse tasks, including inpainting, deraining, and super-resolution. Code is available at https://github.com/Hammour-steak/GOUB.

Distribution matching methods for language model alignment such as Generation with Distributional Control (GDC) and Distributional Policy Gradient (DPG) have not received the same level of attention in reinforcement learning from human feedback (RLHF) as contrastive methods such as Sequence Likelihood Calibration (SLiC), Direct Preference Optimization (DPO) and its variants. We identify high variance of the gradient estimate as the primary reason for the lack of success of these methods and propose a self-normalized baseline to reduce the variance. We further generalize the target distribution in DPG, GDC and DPO by using Bayes' rule to define the reward-conditioned posterior. The resulting approach, referred to as BRAIn - Bayesian Reward-conditioned Amortized Inference acts as a bridge between distribution matching methods and DPO and significantly outperforms prior art in summarization and Antropic HH tasks.

This paper presents a novel approach to object completion, with the primary goal of reconstructing a complete object from its partially visible components. Our method, named MaskComp, delineates the completion process through iterative stages of generation and segmentation. In each iteration, the object mask is provided as an additional condition to boost image generation, and, in return, the generated images can lead to a more accurate mask by fusing the segmentation of images. We demonstrate that the combination of one generation and one segmentation stage effectively functions as a mask denoiser. Through alternation between the generation and segmentation stages, the partial object mask is progressively refined, providing precise shape guidance and yielding superior object completion results. Our experiments demonstrate the superiority of MaskComp over existing approaches, e.g., ControlNet and Stable Diffusion, establishing it as an effective solution for object completion.

This paper enhances image-GPT (iGPT), one of the pioneering works that introduce autoregressive pretraining to predict the next pixels for visual representation learning. Two simple yet essential changes are made. First, we shift the prediction target from raw pixels to semantic tokens, enabling a higher-level understanding of visual content. Second, we supplement the autoregressive modeling by instructing the model to predict not only the next tokens but also the visible tokens. This pipeline is particularly effective when semantic tokens are encoded by discriminatively trained models, such as CLIP. We introduce this novel approach as D-iGPT. Extensive experiments showcase that D-iGPT excels as a strong learner of visual representations: A notable achievement is its compelling performance on the ImageNet-1K dataset --- by training on publicly available datasets, D-iGPT unprecedentedly achieves **90.0%** top-1 accuracy with a vanilla ViT-H. Additionally, D-iGPT shows strong generalization on the downstream task. Code is available at https://github.com/OliverRensu/D-iGPT.

This work targets a novel text-driven **whole-body** motion generation task, which takes a given textual description as input and aims at generating high-quality, diverse, and coherent facial expressions, hand gestures, and body motions simultaneously. Previous works on text-driven motion generation tasks mainly have two limitations: they ignore the key role of fine-grained hand and face controlling in vivid whole-body motion generation, and lack a good alignment between text and motion. To address such limitations, we propose a Text-aligned whOle-body Motion generATiOn framework, named HumanTOMATO, which is the first attempt to our knowledge towards applicable holistic motion generation in this research area. To tackle this challenging task, our solution includes two key designs: (1) a Holistic Hierarchical VQ-VAE (aka H${}^{2}$VQ) and a Hierarchical-GPT for fine-grained body and hand motion reconstruction and generation with two structured codebooks; and (2) a pre-trained text-motion-alignment model to help generated motion align with the input textual description explicitly. Comprehensive experiments verify that our model has significant advantages in both the quality of generated motions and their alignment with text.

One highly promising direction for enabling flexible real-time on-device image editing is utilizing data distillation by leveraging large-scale text-to-image diffusion models to generate paired datasets used for training generative adversarial networks (GANs). This approach notably alleviates the stringent requirements typically imposed by high-end commercial GPUs for performing image editing with diffusion models. However, unlike text-to-image diffusion models, each distilled GAN is specialized for a specific image editing task, necessitating costly training efforts to obtain models for various concepts. In this work, we introduce and address a novel research direction: can the process of distilling GANs from diffusion models be made significantly more efficient? To achieve this goal, we propose a series of innovative techniques. First, we construct a base GAN model with generalized features, adaptable to different concepts through fine-tuning, eliminating the need for training from scratch. Second, we identify crucial layers within the base GAN model and employ Low-Rank Adaptation (LoRA) with a simple yet effective rank search process, rather than fine-tuning the entire base model. Third, we investigate the minimal amount of data necessary for fine-tuning, further reducing the overall training time. Extensive experiments show that we can efficiently empower GANs with the ability to perform real-time high-quality image editing on mobile devices with remarkably reduced training and storage costs for each concept.

We introduce MusicFlow, a cascaded text-to-music generation model based on flow matching. Based on self-supervised representations to bridge between text descriptions and music audios, we construct two flow matching networks to model the conditional distribution of semantic and acoustic features. Additionally, we leverage masked prediction as the training objective, enabling the model to generalize to other tasks such as music infilling and continuation in a zero-shot manner. Experiments on MusicCaps reveal that the music generated by MusicFlow exhibits superior quality and text coherence despite being over $2\sim5$ times smaller and requiring $5$ times fewer iterative steps. Simultaneously, the model can perform other music generation tasks and achieves competitive performance in music infilling and continuation.

In this paper, we present Directly Denoising Diffusion Models (DDDMs): a simple and generic approach for generating realistic images with few-step sampling, while multistep sampling is still preserved for better performance. DDDMs require no delicately designed samplers nor distillation on pre-trained distillation models. DDDMs train the diffusion model conditioned on an estimated target that was generated from previous training iterations of its own. To generate images, samples generated from previous timestep are also taken into consideration, guiding the generation process iteratively. We further propose Pseudo-LPIPS, a novel metric loss that is more robust to various values of hyperparameter. Despite its simplicity, the proposed approach can achieve strong performance in benchmark datasets. Our model achieves FID scores of 2.57 and 2.33 on CIFAR-10 in one-step and two-step sampling respectively, surpassing those obtained from GANs and distillation-based models. By extending the sampling to 1000 steps, we further reduce FID score to 1.79, aligning with state-of-the-art methods in the literature. For ImageNet 64x64, our approach stands as a competitive contender against leading models.

Diffusion models have demonstrated great potential for generating diverse images. However, their performance often suffers from slow generation due to iterative denoising. Knowledge distillation has been recently proposed as a remedy which can reduce the number of inference steps to one or a few, without significant quality degradation. However, existing distillation methods either require significant amounts of offline computation for generating synthetic training data from the teacher model, or need to perform expensive online learning with the help of real data. In this work, we present a novel technique called BOOT, that overcomes these limitations with an efficient data-free distillation algorithm. The core idea is to learn a time-conditioned model that predicts the output of a pre-trained diffusion model teacher given any time-step. Such a model can be efficiently trained based on bootstrapping from two consecutive sampled steps. Furthermore, our method can be easily adapted to large-scale text-to-image diffusion models, which are challenging for previous methods given the fact that the training sets are often large and difficult to access. We demonstrate the effectiveness of our approach on several benchmark datasets in the DDIM setting, achieving comparable generation quality while being orders of magnitude faster than the diffusion teacher. The text-to-image results show that the proposed approach is able to handle highly complex distributions, shedding light on more efficient generative modeling.

Post-hoc explainability methods such as Grad-CAM are popular because they do not influence the performance of a trained model. However, they mainly reveal ''where'' a model looks at for a given input, fail to explain ''what'' the model looks for (e.g., what is important to classify a bird image to a Scott Oriole?). Existing part-prototype networks leverage part-prototypes (e.g., characteristic Scott Oriole's wing and head) to answer both ''where" and ''what", but often under-perform their black box counterparts in the accuracy. Therefore, a natural question is: can one construct a network that answers both ''where'' and ''what" in a post-hoc manner to guarantee the model's performance? To this end, we propose the first post-hoc part-prototype network via decomposing the classification head of a trained model into a set of interpretable part-prototypes. Concretely, we propose an unsupervised prototype discovery and refining strategy to obtain prototypes that can precisely reconstruct the classification head, yet being interpretable. Besides guaranteeing the performance, we show that our network offers more faithful explanations qualitatively and yields even better part-prototypes quantitatively than prior part-prototype networks.

We study the behavior of deterministic methods for solving inverse problems in imaging. These methods are commonly designed to achieve two goals: (1) attaining high perceptual quality, and (2) generating reconstructions that are consistent with the measurements. We provide a rigorous proof that the better a predictor satisfies these two requirements, the larger its Lipschitz constant must be, regardless of the nature of the degradation involved. In particular, to approach perfect perceptual quality and perfect consistency, the Lipschitz constant of the model must grow to infinity. This implies that such methods are necessarily more susceptible to adversarial attacks. We demonstrate our theory on single image super-resolution algorithms, addressing both noisy and noiseless settings. We also show how this undesired behavior can be leveraged to explore the posterior distribution, thereby allowing the deterministic model to imitate stochastic methods.

Due to the growing complexity of modern Integrated Circuits (ICs), there is a need for automated circuit design methods. Recent years have seen increasing research in hardware design language generation to facilitate the design process. In this work, we propose a Verilog generation framework, BetterV, which fine-tunes large language models (LLMs) on processed domain-specific datasets and incorporates generative discriminators for guidance on particular design demands. Verilog modules are collected, filtered, and processed from the internet to form a clean and abundant dataset. Instruct-tuning methods are specially designed to fine-tune the LLMs to understand knowledge about Verilog. Furthermore, data are augmented to enrich the training set and are also used to train a generative discriminator on particular downstream tasks, providing guidance for the LLMs to optimize Verilog implementation. BetterV has the ability to generate syntactically and functionally correct Verilog, outperforming GPT-4 on the VerilogEval benchmark. With the help of task-specific generative discriminators, BetterV achieves remarkable improvements on various electronic design automation (EDA) downstream tasks, including netlist node reduction for synthesis and verification runtime reduction with Boolean Satisfiability (SAT) solving.

We present X-MDPT ($\underline{Cross}$-view $\underline{M}$asked $\underline{D}$iffusion $\underline{P}$rediction $\underline{T}$ransformers), a novel diffusion model designed for pose-guided human image generation. X-MDPT distinguishes itself by employing masked diffusion transformers that operate on latent patches, a departure from the commonly-used Unet structures in existing works. The model comprises three key modules: 1) a denoising diffusion Transformer, 2) an aggregation network that consolidates conditions into a single vector for the diffusion process, and 3) a mask cross-prediction module that enhances representation learning with semantic information from the reference image. X-MDPT demonstrates scalability, improving FID, SSIM, and LPIPS with larger models. Despite its simple design, our model outperforms state-of-the-art approaches on the DeepFashion dataset while exhibiting efficiency in terms of training parameters, training time, and inference speed. Our compact 33MB model achieves an FID of 7.42, surpassing a prior Unet latent diffusion approach (FID 8.07) using only $11\times$ fewer parameters. Our best model surpasses the pixel-based diffusion with $\frac{2}{3}$ of the parameters and achieves $5.43 \times$ faster inference. The code is available at https://github.com/trungpx/xmdpt.