NeurIPS 2024 论文列表

Visual grounding is a common vision task that involves grounding descriptive sentences to the corresponding regions of an image. Most existing methods use independent image-text encoding and apply complex hand-crafted modules or encoder-decoder architectures for modal interaction and query reasoning. However, their performance significantly drops when dealing with complex textual expressions. This is because the former paradigm only utilizes limited downstream data to fit the multi-modal feature fusion. Therefore, it is only effective when the textual expressions are relatively simple. In contrast, given the wide diversity of textual expressions and the uniqueness of downstream training data, the existing fusion module, which extracts multimodal content from a visual-linguistic context, has not been fully investigated. In this paper, we present a simple yet robust transformer-based framework, SimVG, for visual grounding. Specifically, we decouple visual-linguistic feature fusion from downstream tasks by leveraging existing multimodal pre-trained models and incorporating additional object tokens to facilitate deep integration of downstream and pre-training tasks. Furthermore, we design a dynamic weight-balance distillation method in the multi-branch synchronous learning process to enhance the representation capability of the simpler branch. This branch only consists of a lightweight MLP, which simplifies the structure and improves reasoning speed. Experiments on six widely used VG datasets, i.e., RefCOCO/+/g, ReferIt, Flickr30K, and GRefCOCO, demonstrate the superiority of SimVG. Finally, the proposed method not only achieves improvements in efficiency and convergence speed but also attains new state-of-the-art performance on these benchmarks. Codes and models are available at https://github.com/Dmmm1997/SimVG.

Generative text-to-image models enable us to synthesize unlimited amounts of images in a controllable manner, spurring many recent efforts to train vision models with synthetic data. However, every synthetic image ultimately originates from the upstream data used to train the generator. Does the intermediate generator provide additional information over directly training on relevant parts of the upstream data? Grounding this question in the setting of image classification, we compare finetuning on task-relevant, targeted synthetic data generated by Stable Diffusion---a generative model trained on the LAION-2B dataset---against finetuning on targeted real images retrieved directly from LAION-2B. We show that while synthetic data can benefit some downstream tasks, it is universally matched or outperformed by real data from the simple retrieval baseline. Our analysis suggests that this underperformance is partially due to generator artifacts and inaccurate task-relevant visual details in the synthetic images. Overall, we argue that targeted retrieval is a critical baseline to consider when training with synthetic data---a baseline that current methods do not yet surpass. We release code, data, and models at [https://github.com/scottgeng00/unmet-promise/](https://github.com/scottgeng00/unmet-promise).

Video Anomaly Detection (VAD) finds widespread applications in security surveillance, traffic monitoring, industrial monitoring, and healthcare. Despite extensive research efforts, there remains a lack of concise reviews that provide insightful guidance for researchers. Such reviews would serve as quick references to grasp current challenges, research trends, and future directions. In this paper, we present such a review, examining models and datasets from various perspectives. We emphasize the critical relationship between model and dataset, where the quality and diversity of datasets profoundly influence model performance, and dataset development adapts to the evolving needs of emerging approaches. Our review identifies practical issues, including the absence of comprehensive datasets with diverse scenarios. To address this, we introduce a new dataset, Multi-Scenario Anomaly Detection (MSAD), comprising 14 distinct scenarios captured from various camera views. Our dataset has diverse motion patterns and challenging variations, such as different lighting and weather conditions, providing a robust foundation for training superior models. We conduct an in-depth analysis of recent representative models using MSAD and highlight its potential in addressing the challenges of detecting anomalies across diverse and evolving surveillance scenarios.

Hierarchical clustering has usually been addressed by discrete optimization using heuristics or continuous optimization of relaxed scores for hierarchies. In this work, we propose to optimize expected scores under a probabilistic model over hierarchies. (1) We show theoretically that the global optimal values of the expected Dasgupta cost and Tree-Sampling divergence (TSD), two unsupervised metrics for hierarchical clustering, are equal to the optimal values of their discrete counterparts contrary to some relaxed scores. (2) We propose Expected Probabilistic Hierarchies (EPH), a probabilistic model to learn hierarchies in data by optimizing expected scores. EPH uses differentiable hierarchy sampling enabling end-to-end gradient descent based optimization, and an unbiased subgraph sampling approach to scale to large datasets. (3) We evaluate EPH on synthetic and real-world datasets including vector and graph datasets. EPH outperforms all other approaches quantitatively and provides meaningful hierarchies in qualitative evaluations.

3D Gaussian splatting (3DGS) has shown promising results in image rendering and surface reconstruction. However, its potential in volumetric reconstruction tasks, such as X-ray computed tomography, remains under-explored. This paper introduces R$^2$-Gaussian, the first 3DGS-based framework for sparse-view tomographic reconstruction. By carefully deriving X-ray rasterization functions, we discover a previously unknown \emph{integration bias} in the standard 3DGS formulation, which hampers accurate volume retrieval. To address this issue, we propose a novel rectification technique via refactoring the projection from 3D to 2D Gaussians. Our new method presents three key innovations: (1) introducing tailored Gaussian kernels, (2) extending rasterization to X-ray imaging, and (3) developing a CUDA-based differentiable voxelizer. Experiments on synthetic and real-world datasets demonstrate that our method outperforms state-of-the-art approaches in accuracy and efficiency. Crucially, it delivers high-quality results in 4 minutes, which is 12$\times$ faster than NeRF-based methods and on par with traditional algorithms.

Diffusion models have demonstrated competitive performance in missing data imputation (MDI) task. However, directly applying diffusion models to MDI produces suboptimal performance due to two primary defects. First, the sample diversity promoted by diffusion models hinders the accurate inference of missing values. Second, data masking reduces observable indices for model training, obstructing imputation performance. To address these challenges, we introduce $\underline{\text{N}}$egative $\underline{\text{E}}$ntropy-regularized $\underline{\text{W}}$asserstein gradient flow for $\underline{\text{Imp}}$utation (NewImp), enhancing diffusion models for MDI from a gradient flow perspective. To handle the first defect, we incorporate a negative entropy regularization term into the cost functional to suppress diversity and improve accuracy. To handle the second defect, we demonstrate that the imputation procedure of NewImp, induced by the conditional distribution-related cost functional, can equivalently be replaced by that induced by the joint distribution, thereby naturally eliminating the need for data masking. Extensive experiments validate the effectiveness of our method. Code is available at [https://github.com/JustusvLiebig/NewImp](https://github.com/JustusvLiebig/NewImp).

This paper presents MaVEn, an innovative Multi-granularity Visual Encoding framework designed to enhance the capabilities of Multimodal Large Language Models (MLLMs) in multi-image reasoning. Current MLLMs primarily focus on single-image visual understanding, limiting their ability to interpret and integrate information across multiple images. MaVEn addresses this limitation by combining discrete visual symbol sequences, which abstract coarse-grained semantic concepts, with traditional continuous representation sequences that model fine-grained features. This dual approach bridges the semantic gap between visual and textual data, thereby improving the model's ability to process and interpret information from multiple images effectively. Additionally, we design a dynamic reduction mechanism by for long-sequence continuous features to enhance multi-image processing efficiency. Experimental results demonstrate that MaVEn significantly enhances MLLMs' understanding in complex multi-image scenarios, while also improving performance in single-image contexts.

Recent work on pruning large language models (LLMs) has shown that one can eliminate a large number of parameters without compromising performance, making pruning a promising strategy to reduce LLM model size. Existing LLM pruning strategies typically assign uniform pruning ratios across layers, limiting overall pruning ability; and recent work on layerwise pruning of LLMs is often based on heuristics that can easily lead to suboptimal performance. In this paper, we leverage Heavy-Tailed Self-Regularization (HT-SR) Theory, in particular the shape of empirical spectral densities (ESDs) of weight matrices, to design improved layerwise pruning ratios for LLMs. Our analysis reveals a wide variability in how well-trained, and thus relatedly how prunable, different layers of an LLM are. Based on this, we propose AlphaPruning, which uses shape metrics to allocate layerwise sparsity ratios in a more theoretically-principled manner. AlphaPruning can be used in conjunction with multiple existing LLM pruning methods. Our empirical results show that AlphaPruning prunes LLaMA-7B to 80% sparsity while maintaining reasonable perplexity, marking a first in the literature on LLMs.

Despite the remarkable success of transformer-based models in various real-world tasks, their underlying mechanisms remain poorly understood. Recent studies have suggested that transformers can implement gradient descent as an in-context learner for linear regression problems and have developed various theoretical analyses accordingly. However, these works mostly focus on the expressive power of transformers by designing specific parameter constructions, lacking a comprehensive understanding of their inherent working mechanisms post-training. In this study, we consider a sparse linear regression problem and investigate how a trained multi-head transformer performs in-context learning. We experimentally discover that the utilization of multi-heads exhibits different patterns across layers: multiple heads are utilized and essential in the first layer, while usually only a single head is sufficient for subsequent layers. We provide a theoretical explanation for this observation: the first layer preprocesses the context data, and the following layers execute simple optimization steps based on the preprocessed context. Moreover, we demonstrate that such a preprocess-then-optimize algorithm can significantly outperform naive gradient descent and ridge regression algorithms. Further experimental results support our explanations. Our findings offer insights into the benefits of multi-head attention and contribute to understanding the more intricate mechanisms hidden within trained transformers.

Adversarial robustness and privacy of deep learning (DL) models are two widely studied topics in AI security. Adversarial training (AT) is an effective approach to improve the robustness of DL models against adversarial attacks. However, while models with AT demonstrate enhanced robustness, they become more susceptible to membership inference attacks (MIAs), thus increasing the risk of privacy leakage. This indicates a negative trade-off between adversarial robustness and privacy in general deep learning models. Visual prompting is a novel model reprogramming (MR) technique used for fine-tuning pre-trained models, achieving good performance in vision tasks, especially when combined with the label mapping technique. However, the performance of label-mapping-based visual prompting (LM-VP) under adversarial attacks and MIAs lacks evaluation. In this work, we regard the MR of LM-VP as a unified entity, referred to as the LM-VP model, and take a step toward jointly evaluating the adversarial robustness and privacy of LM-VP models. Experimental results show that the choice of pre-trained models significantly affects the white-box adversarial robustness of LM-VP, and standard AT even substantially degrades its performance. In contrast, transfer AT-trained LM-VP achieves a good trade-off between transferred adversarial robustness and privacy, a finding that has been consistently validated across various pre-trained models.

In long-term time series forecasting (LTSF) tasks, an increasing number of works have acknowledged that discrete time series originate from continuous dynamic systems and have attempted to model their underlying dynamics. Recognizing the chaotic nature of real-world data, our model, Attraos, incorporates chaos theory into LTSF, perceiving real-world time series as low-dimensional observations from unknown high-dimensional chaotic dynamical systems. Under the concept of attractor invariance, Attraos utilizes non-parametric Phase Space Reconstruction embedding along with a novel multi-resolution dynamic memory unit to memorize historical dynamical structures, and evolves by a frequency-enhanced local evolution strategy. Detailed theoretical analysis and abundant empirical evidence consistently show that Attraos outperforms various LTSF methods on mainstream LTSF datasets and chaotic datasets with only one-twelfth of the parameters compared to PatchTST.

Matching objectives underpin the success of modern generative models and rely on constructing conditional paths that transform a source distribution into a target distribution. Despite being a fundamental building block, conditional paths have been designed principally under the assumption of $\textit{Euclidean geometry}$, resulting in straight interpolations. However, this can be particularly restrictive for tasks such as trajectory inference, where straight paths might lie outside the data manifold, thus failing to capture the underlying dynamics giving rise to the observed marginals. In this paper, we propose Metric Flow Matching (MFM), a novel simulation-free framework for conditional flow matching where interpolants are approximate geodesics learned by minimizing the kinetic energy of a data-induced Riemannian metric. This way, the generative model matches vector fields on the data manifold, which corresponds to lower uncertainty and more meaningful interpolations. We prescribe general metrics to instantiate MFM, independent of the task, and test it on a suite of challenging problems including LiDAR navigation, unpaired image translation, and modeling cellular dynamics. We observe that MFM outperforms the Euclidean baselines, particularly achieving SOTA on single-cell trajectory prediction.

Neural networks are most often trained under the assumption that data come from a stationary distribution. However, settings in which this assumption is violated are of increasing importance; examples include supervised learning with distributional shifts, reinforcement learning, continual learning and non-stationary contextual bandits. Here, we introduce a novel learning approach that automatically models and adapts to non-stationarity by linking parameters through an Ornstein-Uhlenbeck process with an adaptive drift parameter. The adaptive drift draws the parameters towards the distribution used at initialisation, so the approach can be understood as a form of soft parameter reset. We show empirically that our approach performs well in non-stationary supervised, and off-policy reinforcement learning settings.

Trajectory prediction is a crucial technology to help systems avoid traffic accidents, ensuring safe autonomous driving. Previous methods typically use a fixed-length and sufficiently long trajectory of an agent as observations to predict its future trajectory. However, in real-world scenarios, we often lack the time to gather enough trajectory points before making predictions, e.g., when a car suddenly appears due to an obstruction, the system must make immediate predictions to prevent a collision. This poses a new challenge for trajectory prediction systems, requiring them to be capable of making accurate predictions based on observed trajectories of arbitrary lengths, leading to the failure of existing methods. In this paper, we propose a Length-agnostic Knowledge Distillation framework, named LaKD, which can make accurate trajectory predictions, regardless of the length of observed data. Specifically, considering the fact that long trajectories, containing richer temporal information but potentially additional interference, may perform better or worse than short trajectories, we devise a dynamic length-agnostic knowledge distillation mechanism for exchanging information among trajectories of arbitrary lengths, dynamically determining the transfer direction based on prediction performance. In contrast to traditional knowledge distillation, LaKD employs a unique model that simultaneously serves as both the teacher and the student, potentially causing knowledge collision during the distillation process. Therefore, we design a dynamic soft-masking mechanism, where we first calculate the importance of neuron units and then apply soft-masking to them, so as to safeguard critical units from disruption during the knowledge distillation process. In essence, LaKD is a general and principled framework that can be naturally compatible with existing trajectory prediction models of different architectures. Extensive experiments on three benchmark datasets, Argoverse 1, nuScenes and Argoverse 2, demonstrate the effectiveness of our approach.

Modern deep learning models often exhibit overconfident predictions, inadequately capturing uncertainty. During model optimization, the expected calibration error tends to overfit earlier than classification accuracy, indicating distinct optimization objectives for classification error and calibration error. To ensure consistent optimization of both model accuracy and model calibration, we propose a novel method incorporating a probability-dependent gradient decay coefficient into loss function. This coefficient exhibits a strong correlation with the overall confidence level. To maintain model calibration during optimization, we utilize a proportional-integral-derivative (PID) controller to dynamically adjust this gradient decay rate, where the adjustment relies on the proposed relative calibration error feedback in each epoch, thereby preventing the model from exhibiting over-confidence or under-confidence. Within the PID control system framework, the proposed relative calibration error serves as the control system output, providing an indication of the overall confidence level, while the gradient decay rate functions as the controlled variable. Moreover, recognizing the impact of gradient amplitude of adaptive decay rates, we implement an adaptive learning rate mechanism for gradient compensation to prevent inadequate learning of over-small or over-large gradient. Empirical experiments validate the efficacy of our PID-based adaptive gradient decay rate approach, ensuring consistent optimization of model calibration and model accuracy.

The optimization of the latents and parameters of diffusion models with respect to some differentiable metric defined on the output of the model is a challenging and complex problem. The sampling for diffusion models is done by solving either the *probability flow* ODE or diffusion SDE wherein a neural network approximates the score function allowing a numerical ODE/SDE solver to be used. However, naive backpropagation techniques are memory intensive, requiring the storage of all intermediate states, and face additional complexity in handling the injected noise from the diffusion term of the diffusion SDE. We propose a novel family of bespoke ODE solvers to the continuous adjoint equations for diffusion models, which we call *AdjointDEIS*. We exploit the unique construction of diffusion SDEs to further simplify the formulation of the continuous adjoint equations using *exponential integrators*. Moreover, we provide convergence order guarantees for our bespoke solvers. Significantly, we show that continuous adjoint equations for diffusion SDEs actually simplify to a simple ODE. Lastly, we demonstrate the effectiveness of AdjointDEIS for guided generation with an adversarial attack in the form of the face morphing problem. Our code will be released on our project page [https://zblasingame.github.io/AdjointDEIS/](https://zblasingame.github.io/AdjointDEIS/)

3D generation methods have shown visually compelling results powered by diffusion image priors. However, they often fail to produce realistic geometric details, resulting in overly smooth surfaces or geometric details inaccurately baked in albedo maps. To address this, we introduce a new method that incorporates touch as an additional modality to improve the geometric details of generated 3D assets. We design a lightweight 3D texture field to synthesize visual and tactile textures, guided by 2D diffusion model priors on both visual and tactile domains. We condition the visual texture generation on high-resolution tactile normals and guide the patch-based tactile texture refinement with a customized TextureDreambooth. We further present a multi-part generation pipeline that enables us to synthesize different textures across various regions. To our knowledge, we are the first to leverage high-resolution tactile sensing to enhance geometric details for 3D generation tasks. We evaluate our method in both text-to-3D and image-to-3D settings. Our experiments demonstrate that our method provides customized and realistic fine geometric textures while maintaining accurate alignment between two modalities of vision and touch.

Existing state-of-the-art dense object detection techniques tend to produce a large number of false positive detections on difficult images with complex scenes because they focus on ensuring a high recall. To improve the detection accuracy, we propose an Adaptive Important Region Selection (AIRS) framework guided by Evidential Q-learning coupled with a uniquely designed reward function. Inspired by human visual attention, our detection model conducts object search in a top-down, hierarchical fashion. It starts from the top of the hierarchy with the coarsest granularity and then identifies the potential patches likely to contain objects of interest. It then discards non-informative patches and progressively moves downward on the selected ones for a fine-grained search. The proposed evidential Q-learning systematically encodes epistemic uncertainty in its evidential-Q value to encourage the exploration of unknown patches, especially in the early phase of model training. In this way, the proposed model dynamically balances exploration-exploitation to cover both highly valuable and informative patches. Theoretical analysis and extensive experiments on multiple datasets demonstrate that our proposed framework outperforms the SOTA models.

The nadir objective vector plays a key role in solving multi-objective optimization problems (MOPs), where it is often used to normalize the objective space and guide the search. The current methods for estimating the nadir objective vector perform effectively only on specific MOPs. This paper reveals the limitations of these methods: exact methods can only work on discrete MOPs, while heuristic methods cannot deal with the MOP with a complicated feasible objective region. To fill this gap, we propose a general and rigorous method, namely boundary decomposition for nadir objective vector estimation (BDNE). BDNE scalarizes the MOP into a set of boundary subproblems. By utilizing bilevel optimization, boundary subproblems are optimized and adjusted alternately, thereby refining their optimal solutions to align with the nadir objective vector. We prove that the bilevel optimization identifies the nadir objective vector under mild conditions. We compare BDNE with existing methods on various black-box MOPs. The results conform to the theoretical analysis and show the significant potential of BDNE for real-world application.