Search

Task-aware navigation continues to be a challenging area of research, especially in scenarios involving open vocabulary. Previous studies primarily focus on finding suitable locations for task completion, often overlooking the importance of the robot's pose. However, the robot's orientation is crucial for successfully completing tasks because of how objects are arranged (e.g., to open a refrigerator door). Humans intuitively navigate to objects with the right orientation using semantics and common sense. For instance, when opening a refrigerator, we naturally stand in front of it rather than to the side. Recent advances suggest that Vision-Language Models (VLMs) can provide robots with similar common sense. Therefore, we develop a VLM-driven method called Navigation-to-Gaze (Navi2Gaze) for efficient navigation and object gazing based on task descriptions. This method uses the VLM to score and select the best pose from numerous candidates automatically. In evaluations on multiple photorealistic simulation benchmarks, Navi2Gaze significantly outperforms existing approaches and precisely determines the optimal orientation relative to target objects.

Drawing upon recent advances in language model alignment, we formulate offline Reinforcement Learning as a two-stage optimization problem: First pretraining expressive generative policies on reward-free behavior datasets, then fine-tuning these policies to align with task-specific annotations like Q-values. This strategy allows us to leverage abundant and diverse behavior data to enhance generalization and enable rapid adaptation to downstream tasks using minimal annotations. In particular, we introduce Efficient Diffusion Alignment (EDA) for solving continuous control problems. EDA utilizes diffusion models for behavior modeling. However, unlike previous approaches, we represent diffusion policies as the derivative of a scalar neural network with respect to action inputs. This representation is critical because it enables direct density calculation for diffusion models, making them compatible with existing LLM alignment theories. During policy fine-tuning, we extend preference-based alignment methods like Direct Preference Optimization (DPO) to align diffusion behaviors with continuous Q-functions. Our evaluation on the D4RL benchmark shows that EDA exceeds all baseline methods in overall performance. Notably, EDA maintains about 95\% of performance and still outperforms several baselines given only 1\% of Q-labelled data during fine-tuning.

The recent development of Sora leads to a new era in text-to-video (T2V) generation. Along with this comes the rising concern about its security risks. The generated videos may contain illegal or unethical content, and there is a lack of comprehensive quantitative understanding of their safety, posing a challenge to their reliability and practical deployment. Previous evaluations primarily focus on the quality of video generation. While some evaluations of text-to-image models have considered safety, they cover fewer aspects and do not address the unique temporal risk inherent in video generation. To bridge this research gap, we introduce T2VSafetyBench, a new benchmark designed for conducting safety-critical assessments of text-to-video models. We define 12 critical aspects of video generation safety and construct a malicious prompt dataset using LLMs and jailbreaking prompt attacks. Based on our evaluation results, we draw several important findings, including: 1) no single model excels in all aspects, with different models showing various strengths; 2) the correlation between GPT-4 assessments and manual reviews is generally high; 3) there is a trade-off between the usability and safety of text-to-video generative models. This indicates that as the field of video generation rapidly advances, safety risks are set to surge, highlighting the urgency of prioritizing video safety. We hope that T2VSafetyBench can provide insights for better understanding the safety of video generation in the era of generative AI.

The deployment of pre-trained models (PTMs) has greatly advanced the field of continual learning (CL), enabling positive knowledge transfer and resilience to catastrophic forgetting. To sustain these advantages for sequentially arriving tasks, a promising direction involves keeping the pre-trained backbone frozen while employing parameter-efficient tuning (PET) techniques to instruct representation learning. Despite the popularity of Prompt-based PET for CL, its empirical design often leads to sub-optimal performance in our evaluation of different PTMs and target tasks. To this end, we propose a unified framework for CL with PTMs and PET that provides both theoretical and empirical advancements. We first perform an in-depth theoretical analysis of the CL objective in a pre-training context, decomposing it into hierarchical components namely within-task prediction, task-identity inference and task-adaptive prediction. We then present Hierarchical Decomposition PET (HiDe-PET), an innovative approach that explicitly optimizes the decomposed objective through incorporating task-specific and task-shared knowledge via mainstream PET techniques along with efficient recovery of pre-trained representations. Leveraging this framework, we delve into the distinct impacts of implementation strategy, PET technique and PET architecture, as well as adaptive knowledge accumulation amidst pronounced distribution changes. Finally, across various CL scenarios, our approach demonstrates remarkably superior performance over a broad spectrum of recent strong baselines., Comment: This is a generalized version of our HiDe-Prompt (NeurIPS 2023, Spotlight)

Diffusion models have obtained substantial progress in image-to-video (I2V) generation. However, such models are not fully understood. In this paper, we report a significant but previously overlooked issue in I2V diffusion models (I2V-DMs), namely, conditional image leakage. I2V-DMs tend to over-rely on the conditional image at large time steps, neglecting the crucial task of predicting the clean video from noisy inputs, which results in videos lacking dynamic and vivid motion. We further address this challenge from both inference and training aspects by presenting plug-and-play strategies accordingly. First, we introduce a training-free inference strategy that starts the generation process from an earlier time step to avoid the unreliable late-time steps of I2V-DMs, as well as an initial noise distribution with optimal analytic expressions (Analytic-Init) by minimizing the KL divergence between it and the actual marginal distribution to effectively bridge the training-inference gap. Second, to mitigate conditional image leakage during training, we design a time-dependent noise distribution for the conditional image, which favors high noise levels at large time steps to sufficiently interfere with the conditional image. We validate these strategies on various I2V-DMs using our collected open-domain image benchmark and the UCF101 dataset. Extensive results demonstrate that our methods outperform baselines by producing videos with more dynamic and natural motion without compromising image alignment and temporal consistency. The project page: \url{https://cond-image-leak.github.io/}., Comment: Project page: https://cond-image-leak.github.io/

Communication is essential for advancing science and technology. Quantum communication, in particular, benefits from the use of catalysts. During the communication process, these catalysts enhance performance while remaining unchanged. Although chemical catalysts that undergo deactivation typically perform worse than those that remain unaffected, quantum catalysts, referred to as embezzling catalysts, can surprisingly outperform their non-deactivating counterparts despite experiencing slight alterations. In this work, we employ embezzling quantum catalysts to enhance the transmission of both quantum and classical information. Our results reveal that using embezzling catalysts augments the efficiency of information transmission across noisy quantum channels, ensuring a non-zero catalytic channel capacity. Furthermore, we introduce catalytic superdense coding, demonstrating how embezzling catalysts can enhance the transmission of classical information. Finally, we explore methods to reduce the dimensionality of catalysts, a step toward making quantum catalysis a practical reality., Comment: 16 pages, 9 figures. Comments welcome!

The goal of data attribution for text-to-image models is to identify the training images that most influence the generation of a new image. We can define "influence" by saying that, for a given output, if a model is retrained from scratch without that output's most influential images, the model should then fail to generate that output image. Unfortunately, directly searching for these influential images is computationally infeasible, since it would require repeatedly retraining from scratch. We propose a new approach that efficiently identifies highly-influential images. Specifically, we simulate unlearning the synthesized image, proposing a method to increase the training loss on the output image, without catastrophic forgetting of other, unrelated concepts. Then, we find training images that are forgotten by proxy, identifying ones with significant loss deviations after the unlearning process, and label these as influential. We evaluate our method with a computationally intensive but "gold-standard" retraining from scratch and demonstrate our method's advantages over previous methods., Comment: Project page: https://peterwang512.github.io/AttributeByUnlearning Code: https://github.com/PeterWang512/AttributeByUnlearning

Despite the superior capabilities of Multimodal Large Language Models (MLLMs) across diverse tasks, they still face significant trustworthiness challenges. Yet, current literature on the assessment of trustworthy MLLMs remains limited, lacking a holistic evaluation to offer thorough insights into future improvements. In this work, we establish MultiTrust, the first comprehensive and unified benchmark on the trustworthiness of MLLMs across five primary aspects: truthfulness, safety, robustness, fairness, and privacy. Our benchmark employs a rigorous evaluation strategy that addresses both multimodal risks and cross-modal impacts, encompassing 32 diverse tasks with self-curated datasets. Extensive experiments with 21 modern MLLMs reveal some previously unexplored trustworthiness issues and risks, highlighting the complexities introduced by the multimodality and underscoring the necessity for advanced methodologies to enhance their reliability. For instance, typical proprietary models still struggle with the perception of visually confusing images and are vulnerable to multimodal jailbreaking and adversarial attacks; MLLMs are more inclined to disclose privacy in text and reveal ideological and cultural biases even when paired with irrelevant images in inference, indicating that the multimodality amplifies the internal risks from base LLMs. Additionally, we release a scalable toolbox for standardized trustworthiness research, aiming to facilitate future advancements in this important field. Code and resources are publicly available at: https://multi-trust.github.io/., Comment: 100 pages, 84 figures, 33 tables

Creating 3D assets from single-view images is a complex task that demands a deep understanding of the world. Recently, feed-forward 3D generative models have made significant progress by training large reconstruction models on extensive 3D datasets, with triplanes being the preferred 3D geometry representation. However, effectively utilizing the geometric priors of triplanes, while minimizing artifacts caused by generated inconsistent multi-view images, remains a challenge. In this work, we present \textbf{Fre}quency modulat\textbf{e}d tri\textbf{plane} (\textbf{Freeplane}), a simple yet effective method to improve the generation quality of feed-forward models without additional training. We first analyze the role of triplanes in feed-forward methods and find that the inconsistent multi-view images introduce high-frequency artifacts on triplanes, leading to low-quality 3D meshes. Based on this observation, we propose strategically filtering triplane features and combining triplanes before and after filtering to produce high-quality textured meshes. These techniques incur no additional cost and can be seamlessly integrated into pre-trained feed-forward models to enhance their robustness against the inconsistency of generated multi-view images. Both qualitative and quantitative results demonstrate that our method improves the performance of feed-forward models by simply modulating triplanes. All you need is to modulate the triplanes during inference., Comment: project can be found in: https://freeplane3d.github.io/

Transformer has shown promise in reinforcement learning to model time-varying features for obtaining generalized low-level robot policies on diverse robotics datasets in embodied learning. However, it still suffers from the issues of low data efficiency and high inference latency. In this paper, we propose to investigate the task from a new perspective of the frequency domain. We first observe that the energy density in the frequency domain of a robot's trajectory is mainly concentrated in the low-frequency part. Then, we present the Fourier Controller Network (FCNet), a new network that uses Short-Time Fourier Transform (STFT) to extract and encode time-varying features through frequency domain interpolation. In order to do real-time decision-making, we further adopt FFT and Sliding DFT methods in the model architecture to achieve parallel training and efficient recurrent inference. Extensive results in both simulated (e.g., D4RL) and real-world environments (e.g., robot locomotion) demonstrate FCNet's substantial efficiency and effectiveness over existing methods such as Transformer, e.g., FCNet outperforms Transformer on multi-environmental robotics datasets of all types of sizes (from 1.9M to 120M). The project page and code can be found https://thkkk.github.io/fcnet.

This paper studies the challenging black-box adversarial attack that aims to generate adversarial examples against a black-box model by only using output feedback of the model to input queries. Some previous methods improve the query efficiency by incorporating the gradient of a surrogate white-box model into query-based attacks due to the adversarial transferability. However, the localized gradient is not informative enough, making these methods still query-intensive. In this paper, we propose a Prior-guided Bayesian Optimization (P-BO) algorithm that leverages the surrogate model as a global function prior in black-box adversarial attacks. As the surrogate model contains rich prior information of the black-box one, P-BO models the attack objective with a Gaussian process whose mean function is initialized as the surrogate model's loss. Our theoretical analysis on the regret bound indicates that the performance of P-BO may be affected by a bad prior. Therefore, we further propose an adaptive integration strategy to automatically adjust a coefficient on the function prior by minimizing the regret bound. Extensive experiments on image classifiers and large vision-language models demonstrate the superiority of the proposed algorithm in reducing queries and improving attack success rates compared with the state-of-the-art black-box attacks. Code is available at https://github.com/yibo-miao/PBO-Attack., Comment: ICML 2024

Deep Neural Networks (DNNs) have achieved remarkable success in a variety of tasks, especially when it comes to prediction accuracy. However, in complex real-world scenarios, particularly in safety-critical applications, high accuracy alone is not enough. Reliable uncertainty estimates are crucial. Modern DNNs, often trained with cross-entropy loss, tend to be overconfident, especially with ambiguous samples. To improve uncertainty calibration, many techniques have been developed, but they often compromise prediction accuracy. To tackle this challenge, we propose the ``mutual-transport ensemble'' (MTE). This approach introduces a co-trained auxiliary model and adaptively regularizes the cross-entropy loss using Kullback-Leibler (KL) divergence between the prediction distributions of the primary and auxiliary models. We conducted extensive studies on various benchmarks to validate the effectiveness of our method. The results show that MTE can simultaneously enhance both accuracy and uncertainty calibration. For example, on the CIFAR-100 dataset, our MTE method on ResNet34/50 achieved significant improvements compared to previous state-of-the-art method, with absolute accuracy increases of 2.4%/3.7%, relative reductions in ECE of $42.3%/29.4%, and relative reductions in classwise-ECE of 11.6%/15.3%.

Generating compact and sharply detailed 3D meshes poses a significant challenge for current 3D generative models. Different from extracting dense meshes from neural representation, some recent works try to model the native mesh distribution (i.e., a set of triangles), which generates more compact results as humans crafted. However, due to the complexity and variety of mesh topology, these methods are typically limited to small datasets with specific categories and are hard to extend. In this paper, we introduce a generic and scalable mesh generation framework PivotMesh, which makes an initial attempt to extend the native mesh generation to large-scale datasets. We employ a transformer-based auto-encoder to encode meshes into discrete tokens and decode them from face level to vertex level hierarchically. Subsequently, to model the complex typology, we first learn to generate pivot vertices as coarse mesh representation and then generate the complete mesh tokens with the same auto-regressive Transformer. This reduces the difficulty compared with directly modeling the mesh distribution and further improves the model controllability. PivotMesh demonstrates its versatility by effectively learning from both small datasets like Shapenet, and large-scale datasets like Objaverse and Objaverse-xl. Extensive experiments indicate that PivotMesh can generate compact and sharp 3D meshes across various categories, highlighting its great potential for native mesh modeling., Comment: Project website: https://whaohan.github.io/pivotmesh

Autoregressively trained transformers have brought a profound revolution to the world, especially with their in-context learning (ICL) ability to address downstream tasks. Recently, several studies suggest that transformers learn a mesa-optimizer during autoregressive (AR) pretraining to implement ICL. Namely, the forward pass of the trained transformer is equivalent to optimizing an inner objective function in-context. However, whether the practical non-convex training dynamics will converge to the ideal mesa-optimizer is still unclear. Towards filling this gap, we investigate the non-convex dynamics of a one-layer linear causal self-attention model autoregressively trained by gradient flow, where the sequences are generated by an AR process $x_{t+1} = W x_t$. First, under a certain condition of data distribution, we prove that an autoregressively trained transformer learns $W$ by implementing one step of gradient descent to minimize an ordinary least squares (OLS) problem in-context. It then applies the learned $\widehat{W}$ for next-token prediction, thereby verifying the mesa-optimization hypothesis. Next, under the same data conditions, we explore the capability limitations of the obtained mesa-optimizer. We show that a stronger assumption related to the moments of data is the sufficient and necessary condition that the learned mesa-optimizer recovers the distribution. Besides, we conduct exploratory analyses beyond the first data condition and prove that generally, the trained transformer will not perform vanilla gradient descent for the OLS problem. Finally, our simulation results verify the theoretical results., Comment: 37pages

Video generative models are receiving particular attention given their ability to generate realistic and imaginative frames. Besides, these models are also observed to exhibit strong 3D consistency, significantly enhancing their potential to act as world simulators. In this work, we present Vidu4D, a novel reconstruction model that excels in accurately reconstructing 4D (i.e., sequential 3D) representations from single generated videos, addressing challenges associated with non-rigidity and frame distortion. This capability is pivotal for creating high-fidelity virtual contents that maintain both spatial and temporal coherence. At the core of Vidu4D is our proposed Dynamic Gaussian Surfels (DGS) technique. DGS optimizes time-varying warping functions to transform Gaussian surfels (surface elements) from a static state to a dynamically warped state. This transformation enables a precise depiction of motion and deformation over time. To preserve the structural integrity of surface-aligned Gaussian surfels, we design the warped-state geometric regularization based on continuous warping fields for estimating normals. Additionally, we learn refinements on rotation and scaling parameters of Gaussian surfels, which greatly alleviates texture flickering during the warping process and enhances the capture of fine-grained appearance details. Vidu4D also contains a novel initialization state that provides a proper start for the warping fields in DGS. Equipping Vidu4D with an existing video generative model, the overall framework demonstrates high-fidelity text-to-4D generation in both appearance and geometry., Comment: Project page: https://vidu4d-dgs.github.io

Denoising diffusion bridge models (DDBMs) are a powerful variant of diffusion models for interpolating between two arbitrary paired distributions given as endpoints. Despite their promising performance in tasks like image translation, DDBMs require a computationally intensive sampling process that involves the simulation of a (stochastic) differential equation through hundreds of network evaluations. In this work, we present diffusion bridge implicit models (DBIMs) for accelerated sampling of diffusion bridges without extra training. We generalize DDBMs via a class of non-Markovian diffusion bridges defined on the discretized timesteps concerning sampling, which share the same training objective as DDBMs. These generalized diffusion bridges give rise to generative processes ranging from stochastic to deterministic (i.e., an implicit probabilistic model) while being up to 25$\times$ faster than the vanilla sampler of DDBMs. Moreover, the deterministic sampling procedure yielded by DBIMs enables faithful encoding and reconstruction by a booting noise used in the initial sampling step, and allows us to perform semantically meaningful interpolation in image translation tasks by regarding the booting noise as the latent variable.

Designing generalizable agents capable of adapting to diverse embodiments has achieved significant attention in Reinforcement Learning (RL), which is critical for deploying RL agents in various real-world applications. Previous Cross-Embodiment RL approaches have focused on transferring knowledge across embodiments within specific tasks. These methods often result in knowledge tightly coupled with those tasks and fail to adequately capture the distinct characteristics of different embodiments. To address this limitation, we introduce the notion of Cross-Embodiment Unsupervised RL (CEURL), which leverages unsupervised learning to enable agents to acquire embodiment-aware and task-agnostic knowledge through online interactions within reward-free environments. We formulate CEURL as a novel Controlled Embodiment Markov Decision Process (CE-MDP) and systematically analyze CEURL's pre-training objectives under CE-MDP. Based on these analyses, we develop a novel algorithm Pre-trained Embodiment-Aware Control (PEAC) for handling CEURL, incorporating an intrinsic reward function specifically designed for cross-embodiment pre-training. PEAC not only provides an intuitive optimization strategy for cross-embodiment pre-training but also can integrate flexibly with existing unsupervised RL methods, facilitating cross-embodiment exploration and skill discovery. Extensive experiments in both simulated (e.g., DMC and Robosuite) and real-world environments (e.g., legged locomotion) demonstrate that PEAC significantly improves adaptation performance and cross-embodiment generalization, demonstrating its effectiveness in overcoming the unique challenges of CEURL.

Despite the successful enhancement to the Harrow-Hassidim-Lloyd algorithm by Childs et al., who introduced the Fourier approach leveraging linear combinations of unitary operators, our research has identified non-trivial redundancies within this method. This finding points to a considerable potential for refinement. In this paper, we propose the quantum Krylov-subspace method (QKSM), which is a hybrid classical-quantum algorithm, to mitigate such redundancies. By integrating QKSM as a subroutine, we introduce the quantum Krylov-subspace method based linear solver that not only reduces computational redundancy but also enhances efficiency and accuracy. Extensive numerical experiments, conducted on systems with dimensions up to $2^{10} \times 2^{10}$, have demonstrated a significant reduction in computational resources and have led to more precise approximations.

We propose a method to distill a complex multistep diffusion model into a single-step conditional GAN student model, dramatically accelerating inference, while preserving image quality. Our approach interprets diffusion distillation as a paired image-to-image translation task, using noise-to-image pairs of the diffusion model's ODE trajectory. For efficient regression loss computation, we propose E-LatentLPIPS, a perceptual loss operating directly in diffusion model's latent space, utilizing an ensemble of augmentations. Furthermore, we adapt a diffusion model to construct a multi-scale discriminator with a text alignment loss to build an effective conditional GAN-based formulation. E-LatentLPIPS converges more efficiently than many existing distillation methods, even accounting for dataset construction costs. We demonstrate that our one-step generator outperforms cutting-edge one-step diffusion distillation models -- DMD, SDXL-Turbo, and SDXL-Lightning -- on the zero-shot COCO benchmark., Comment: Project page: https://mingukkang.github.io/Diffusion2GAN/

We introduce Vidu, a high-performance text-to-video generator that is capable of producing 1080p videos up to 16 seconds in a single generation. Vidu is a diffusion model with U-ViT as its backbone, which unlocks the scalability and the capability for handling long videos. Vidu exhibits strong coherence and dynamism, and is capable of generating both realistic and imaginative videos, as well as understanding some professional photography techniques, on par with Sora -- the most powerful reported text-to-video generator. Finally, we perform initial experiments on other controllable video generation, including canny-to-video generation, video prediction and subject-driven generation, which demonstrate promising results., Comment: Project page at https://www.shengshu-ai.com/vidu

Topological data analysis (TDA) uses persistent homology to quantify loops and higher-dimensional holes in data, making it particularly relevant for examining the characteristics of images of cells in the field of cell biology. In the context of a cell injury, as time progresses, a wound in the form of a ring emerges in the cell image and then gradually vanishes. Performing statistical inference on this ring-like pattern in a single image is challenging due to the absence of repeated samples. In this paper, we develop a novel framework leveraging TDA to estimate underlying structures within individual images and quantify associated uncertainties through confidence regions. Our proposed method partitions the image into the background and the damaged cell regions. Then pixels within the affected cell region are used to establish confidence regions in the space of persistence diagrams (topological summary statistics). The method establishes estimates on the persistence diagrams which correct the bias of traditional TDA approaches. A simulation study is conducted to evaluate the coverage probabilities of the proposed confidence regions in comparison to an alternative approach is proposed in this paper. We also illustrate our methodology by a real-world example provided by cell repair., Comment: 30 pages, 8 figures

Art reinterpretation is the practice of creating a variation of a reference work, making a paired artwork that exhibits a distinct artistic style. We ask if such an image pair can be used to customize a generative model to capture the demonstrated stylistic difference. We propose Pair Customization, a new customization method that learns stylistic difference from a single image pair and then applies the acquired style to the generation process. Unlike existing methods that learn to mimic a single concept from a collection of images, our method captures the stylistic difference between paired images. This allows us to apply a stylistic change without overfitting to the specific image content in the examples. To address this new task, we employ a joint optimization method that explicitly separates the style and content into distinct LoRA weight spaces. We optimize these style and content weights to reproduce the style and content images while encouraging their orthogonality. During inference, we modify the diffusion process via a new style guidance based on our learned weights. Both qualitative and quantitative experiments show that our method can effectively learn style while avoiding overfitting to image content, highlighting the potential of modeling such stylistic differences from a single image pair., Comment: project page: https://paircustomization.github.io/

Optimization-based approaches, such as score distillation sampling (SDS), show promise in zero-shot 3D generation but suffer from low efficiency, primarily due to the high number of function evaluations (NFEs) required for each sample. In this paper, we introduce score-based iterative reconstruction (SIR), an efficient and general algorithm mimicking a differentiable 3D reconstruction process to reduce the NFEs. Given a single set of images sampled from a multi-view score-based diffusion model, SIR repeatedly optimizes 3D parameters, unlike the single-step optimization in SDS. With other improvements in training, we present an efficient approach called MicroDreamer that generally applies to various 3D representations and 3D generation tasks. In particular, retaining a comparable performance, MicroDreamer is 5-20 times faster than SDS in generating neural radiance field and takes about 20 seconds to generate meshes from 3D Gaussian splatting on a single A100 GPU, halving the time of the fastest zero-shot baseline, DreamGaussian. Our code is available at \url{https://github.com/ML-GSAI/MicroDreamer}.

Existing pedestrian attribute recognition (PAR) algorithms are mainly developed based on a static image, however, the performance is unreliable in challenging scenarios, such as heavy occlusion, motion blur, etc. In this work, we propose to understand human attributes using video frames that can fully use temporal information by fine-tuning a pre-trained multi-modal foundation model efficiently. Specifically, we formulate the video-based PAR as a vision-language fusion problem and adopt a pre-trained foundation model CLIP to extract the visual features. More importantly, we propose a novel spatiotemporal side-tuning strategy to achieve parameter-efficient optimization of the pre-trained vision foundation model. To better utilize the semantic information, we take the full attribute list that needs to be recognized as another input and transform the attribute words/phrases into the corresponding sentence via split, expand, and prompt operations. Then, the text encoder of CLIP is utilized for embedding processed attribute descriptions. The averaged visual tokens and text tokens are concatenated and fed into a fusion Transformer for multi-modal interactive learning. The enhanced tokens will be fed into a classification head for pedestrian attribute prediction. Extensive experiments on two large-scale video-based PAR datasets fully validated the effectiveness of our proposed framework. The source code of this paper is available at https://github.com/Event-AHU/OpenPAR., Comment: Parameter Efficient Fine-Tuning Strategy for Video-based Pedestrian Attribute Recognition

Fr\'echet Video Distance (FVD), a prominent metric for evaluating video generation models, is known to conflict with human perception occasionally. In this paper, we aim to explore the extent of FVD's bias toward per-frame quality over temporal realism and identify its sources. We first quantify the FVD's sensitivity to the temporal axis by decoupling the frame and motion quality and find that the FVD increases only slightly with large temporal corruption. We then analyze the generated videos and show that via careful sampling from a large set of generated videos that do not contain motions, one can drastically decrease FVD without improving the temporal quality. Both studies suggest FVD's bias towards the quality of individual frames. We further observe that the bias can be attributed to the features extracted from a supervised video classifier trained on the content-biased dataset. We show that FVD with features extracted from the recent large-scale self-supervised video models is less biased toward image quality. Finally, we revisit a few real-world examples to validate our hypothesis., Comment: CVPR 2024. Project webpage: https://content-debiased-fvd.github.io/

Model customization introduces new concepts to existing text-to-image models, enabling the generation of the new concept in novel contexts. However, such methods lack accurate camera view control w.r.t the object, and users must resort to prompt engineering (e.g., adding "top-view") to achieve coarse view control. In this work, we introduce a new task -- enabling explicit control of camera viewpoint for model customization. This allows us to modify object properties amongst various background scenes via text prompts, all while incorporating the target camera pose as additional control. This new task presents significant challenges in merging a 3D representation from the multi-view images of the new concept with a general, 2D text-to-image model. To bridge this gap, we propose to condition the 2D diffusion process on rendered, view-dependent features of the new object. During training, we jointly adapt the 2D diffusion modules and 3D feature predictions to reconstruct the object's appearance and geometry while reducing overfitting to the input multi-view images. Our method outperforms existing image editing and model personalization baselines in preserving the custom object's identity while following the input text prompt and the object's camera pose., Comment: project page: https://customdiffusion360.github.io

Although Multimodal Large Language Models (MLLMs) have demonstrated promising versatile capabilities, their performance is still inferior to specialized models on downstream tasks, which makes adaptation necessary to enhance their utility. However, fine-tuning methods require independent training for every model, leading to huge computation and memory overheads. In this paper, we propose a novel setting where we aim to improve the performance of diverse MLLMs with a group of shared parameters optimized for a downstream task. To achieve this, we propose Transferable Visual Prompting (TVP), a simple and effective approach to generate visual prompts that can transfer to different models and improve their performance on downstream tasks after trained on only one model. We introduce two strategies to address the issue of cross-model feature corruption of existing visual prompting methods and enhance the transferability of the learned prompts, including 1) Feature Consistency Alignment: which imposes constraints to the prompted feature changes to maintain task-agnostic knowledge; 2) Task Semantics Enrichment: which encourages the prompted images to contain richer task-specific semantics with language guidance. We validate the effectiveness of TVP through extensive experiments with 6 modern MLLMs on a wide variety of tasks ranging from object recognition and counting to multimodal reasoning and hallucination correction., Comment: Accepted in CVPR 2024 as Poster (Highlight)

Diffusion models have been extensively used in data generation tasks and are recognized as one of the best generative models. However, their time-consuming deployment, long inference time, and requirements on large memory limit their application on mobile devices. In this paper, we propose a method based on the improved Straight-Through Estimator to improve the deployment efficiency of diffusion models. Specifically, we add sparse masks to the Convolution and Linear layers in a pre-trained diffusion model, then use design progressive sparsity for model training in the fine-tuning stage, and switch the inference mask on and off, which supports a flexible choice of sparsity during inference according to the FID and MACs requirements. Experiments on four datasets conducted on a state-of-the-art Transformer-based diffusion model demonstrate that our method reduces MACs by $50\%$ while increasing FID by only 1.5 on average. Under other MACs conditions, the FID is also lower than 1$\sim$137 compared to other methods.

Training large transformers is slow, but recent innovations on GPU architecture give us an advantage. NVIDIA Ampere GPUs can execute a fine-grained 2:4 sparse matrix multiplication twice as fast as its dense equivalent. In the light of this property, we comprehensively investigate the feasibility of accelerating feed-forward networks (FFNs) of transformers in pre-training. First, we define a ``flip rate'' to monitor the stability of a 2:4 training process. Utilizing this metric, we propose three techniques to preserve accuracy: to modify the sparse-refined straight-through estimator by applying the masked decay term on gradients, to determine a feasible decay factor in warm-up stage, and to enhance the model's quality by a dense fine-tuning procedure near the end of pre-training. Besides, we devise two techniques to practically accelerate training: to calculate transposable 2:4 masks by convolution, and to accelerate gated activation functions by reducing GPU L2 cache miss. Experiments show that our 2:4 sparse training algorithm achieves similar convergence to dense training algorithms on several transformer pre-training tasks, while actual acceleration can be observed on different shapes of transformer block apparently. Our toolkit is available at https://github.com/huyz2023/2by4-pretrain.

3D content generation has wide applications in various fields. One of its dominant paradigms is by sparse-view reconstruction using multi-view images generated by diffusion models. However, since directly reconstructing triangle meshes from multi-view images is challenging, most methodologies opt to an implicit representation (such as NeRF) during the sparse-view reconstruction and acquire the target mesh by a post-processing extraction. However, the implicit representation takes extensive time to train and the post-extraction also leads to undesirable visual artifacts. In this paper, we propose FlexiDreamer, a novel framework that directly reconstructs high-quality meshes from multi-view generated images. We utilize an advanced gradient-based mesh optimization, namely FlexiCubes, for multi-view mesh reconstruction, which enables us to generate 3D meshes in an end-to-end manner. To address the reconstruction artifacts owing to the inconsistencies from generated images, we design a hybrid positional encoding scheme to improve the reconstruction geometry and an orientation-aware texture mapping to mitigate surface ghosting. To further enhance the results, we respectively incorporate eikonal and smooth regularizations to reduce geometric holes and surface noise. Our approach can generate high-fidelity 3D meshes in the single image-to-3D downstream task with approximately 1 minute, significantly outperforming previous methods., Comment: Project page: https://flexidreamer.github.io

The vulnerability of deep neural networks to adversarial patches has motivated numerous defense strategies for boosting model robustness. However, the prevailing defenses depend on single observation or pre-established adversary information to counter adversarial patches, often failing to be confronted with unseen or adaptive adversarial attacks and easily exhibiting unsatisfying performance in dynamic 3D environments. Inspired by active human perception and recurrent feedback mechanisms, we develop Embodied Active Defense (EAD), a proactive defensive strategy that actively contextualizes environmental information to address misaligned adversarial patches in 3D real-world settings. To achieve this, EAD develops two central recurrent sub-modules, i.e., a perception module and a policy module, to implement two critical functions of active vision. These models recurrently process a series of beliefs and observations, facilitating progressive refinement of their comprehension of the target object and enabling the development of strategic actions to counter adversarial patches in 3D environments. To optimize learning efficiency, we incorporate a differentiable approximation of environmental dynamics and deploy patches that are agnostic to the adversary strategies. Extensive experiments demonstrate that EAD substantially enhances robustness against a variety of patches within just a few steps through its action policy in safety-critical tasks (e.g., face recognition and object detection), without compromising standard accuracy. Furthermore, due to the attack-agnostic characteristic, EAD facilitates excellent generalization to unseen attacks, diminishing the averaged attack success rate by 95 percent across a range of unseen adversarial attacks., Comment: 27pages

3D content creation from text prompts has shown remarkable success recently. However, current text-to-3D methods often generate 3D results that do not align well with human preferences. In this paper, we present a comprehensive framework, coined DreamReward, to learn and improve text-to-3D models from human preference feedback. To begin with, we collect 25k expert comparisons based on a systematic annotation pipeline including rating and ranking. Then, we build Reward3D -- the first general-purpose text-to-3D human preference reward model to effectively encode human preferences. Building upon the 3D reward model, we finally perform theoretical analysis and present the Reward3D Feedback Learning (DreamFL), a direct tuning algorithm to optimize the multi-view diffusion models with a redefined scorer. Grounded by theoretical proof and extensive experiment comparisons, our DreamReward successfully generates high-fidelity and 3D consistent results with significant boosts in prompt alignment with human intention. Our results demonstrate the great potential for learning from human feedback to improve text-to-3D models., Comment: Project page: https://jamesyjl.github.io/DreamReward

Pretraining transformers are generally time-consuming. Fully quantized training (FQT) is a promising approach to speed up pretraining. However, most FQT methods adopt a quantize-compute-dequantize procedure, which often leads to suboptimal speedup and significant performance degradation when used in transformers due to the high memory access overheads and low-precision computations. In this work, we propose Jetfire, an efficient and accurate INT8 training method specific to transformers. Our method features an INT8 data flow to optimize memory access and a per-block quantization method to maintain the accuracy of pretrained transformers. Extensive experiments demonstrate that our INT8 FQT method achieves comparable accuracy to the FP16 training baseline and outperforms the existing INT8 training works for transformers. Moreover, for a standard transformer block, our method offers an end-to-end training speedup of 1.42x and a 1.49x memory reduction compared to the FP16 baseline., Comment: 14 pages, 8 figures

In this work, we address two limitations of existing conditional diffusion models: their slow inference speed due to the iterative denoising process and their reliance on paired data for model fine-tuning. To tackle these issues, we introduce a general method for adapting a single-step diffusion model to new tasks and domains through adversarial learning objectives. Specifically, we consolidate various modules of the vanilla latent diffusion model into a single end-to-end generator network with small trainable weights, enhancing its ability to preserve the input image structure while reducing overfitting. We demonstrate that, for unpaired settings, our model CycleGAN-Turbo outperforms existing GAN-based and diffusion-based methods for various scene translation tasks, such as day-to-night conversion and adding/removing weather effects like fog, snow, and rain. We extend our method to paired settings, where our model pix2pix-Turbo is on par with recent works like Control-Net for Sketch2Photo and Edge2Image, but with a single-step inference. This work suggests that single-step diffusion models can serve as strong backbones for a range of GAN learning objectives. Our code and models are available at https://github.com/GaParmar/img2img-turbo., Comment: Github: https://github.com/GaParmar/img2img-turbo