Contrastive Language-Image Pre-training (CLIP) [37] has emerged as a pivotal model in computer vision and multimodal learning, achieving state-of-the-art performance at aligning visual and textual representations through contrastive learning. However, CLIP struggles with potential information misalignment in many image-text datasets and suffers from entangled representation. On the one hand, short captions for a single image in datasets like MSCOCO may describe disjoint regions in the image, leaving the model uncertain about which visual features to retain or disregard. On the other hand, directly aligning long captions with images can lead to the retention of entangled details, preventing the model from learning disentangled, atomic concepts – ultimately limiting its generalization on certain downstream tasks involving short prompts.In this paper, we establish theoretical conditions that enable flexible alignment between textual and visual representations across varying levels of granularity. Specifically, our framework ensures that a model can not only preserve cross-modal semantic information in its entirety but also disentangle visual representations to capture finegrained textual concepts. Building on this foundation, we introduce SmartCLIP, a novel approach that identifies and aligns the most relevant visual and textual representations in a modular manner. Superior performance across various tasks demonstrates its capability to handle information misalignment and supports our identification theory. The code is available at https://github.com/MidPush/SmartCLIP.

Multi-agent frameworks powered by large language models (LLMs) have demonstrated great success in automated planning and task execution. However, the effective adjustment of agentic workflows during execution has not been well studied. An effective workflow adjustment is crucial in real-world scenarios, as the initial plan must adjust to unforeseen challenges and changing conditions in real time to ensure the efficient execution of complex tasks. In this paper, we define workflows as an activity-on-vertex (AOV) graph, which allows continuous workflow refinement by LLM agents through dynamic subtask allocation adjustment based on historical performance and previous AOVs. To further enhance framework performance, we emphasize modularity in workflow design based on evaluating parallelism and dependency complexity. With this design, our proposed multi-agent framework achieves efficient concurrent execution of subtasks, effective goal achievement, and enhanced error tolerance. Empirical results across various practical tasks demonstrate significant improvements in the efficiency of multi-agent frameworks through dynamic workflow refinement and modularization.

Understanding whether the data generative process follows causal or anticausal relations is important for many applications. Existing causal discovery methods struggle with high-dimensional perceptual data such as images. Moreover, they require well-labeled data, which may not be feasible due to measurement error. In this paper, we propose a robust method to detect whether the data generative process is causal or anticausal. To determine the causal or anticausal relation, we identify an asymmetric property: under the causal relation, the instance distribution does not contain information about the noisy class-posterior distribution. We also propose a practical method to verify this via a noise injection approach. Our method is robust to label errors and is designed to handle both large-scale and high-dimensional datasets effectively. Both theoretical analyses and empirical results on a variety of datasets demonstrate the effectiveness of our proposed method in determining the causal or anticausal direction of the data generative process.

In-context learning (ICL) has revolutionized natural language processing by enabling models to adapt to diverse tasks with only a few illustrative examples. However, the exploration of ICL within the field of computer vision remains limited. Inspired by Chain-of-Thought (CoT) prompting in the language domain, we propose Chain-of-Focus (CoF) Prompting, which enhances vision models by enabling step-by-step visual comprehension. CoF Prompting addresses the challenges of absent logical structure in visual data by generating intermediate reasoning steps through visual saliency. Moreover, it provides a solution for creating tailored prompts from visual inputs by selecting contextually informative prompts based on query similarity and target richness. The significance of CoF prompting is demonstrated by the recent introduction of Large Autoregressive Vision Models (LAVMs), which predict downstream targets via in-context learning with pure visual inputs. By integrating intermediate reasoning steps into visual prompts and effectively selecting the informative ones, the LAVMs are capable of generating significantly better inferences. Extensive experiments on downstream visual understanding tasks validate the effectiveness of our proposed method for visual in-context learning.

Semantic Dependency refers to the relationship between words in a sentence where the meaning of one word depends on another, which is important for natural language understanding. In this paper, we investigate the role of semantic dependencies in answering questions for transformer models, which is achieved by analyzing how token values shift in response to changes in semantics. Through extensive experiments on models including the BERT series, GPT, and LLaMA, we uncover the following key findings: 1). Most tokens primarily retain their original semantic information even as they propagate through multiple layers. 2). Models can encode truthful semantic dependencies in tokens in the final layer. 3). Mistakes in model answers often stem from specific tokens encoded with incorrect semantic dependencies. Furthermore, we found that addressing the incorrectness by directly adjusting parameters is challenging because the same parameters can encode both correct and incorrect semantic dependencies depending on the context. Our findings provide insights into the causes of incorrect information generation in transformers and help the future development of robust and reliable models.

Conditional independence (CI) test is a fundamental concept in statistics. In many real-world scenarios, some variables may be difficult to measure accurately, often leading to data being represented as discretized values. Applying CI tests directly to discretized data, however, can lead to incorrect conclusions about the independence of latent variables. To address this, recent advancements have sought to infer the correct CI relationship between the latent variables by binarizing the observed data. However, this process results in a loss of information, which degrades the test’s performance, particularly with small sample sizes. Motivated by this, this paper introduces a new sample-efficient CI test that does not rely on the binarization process. We find that the relationship can be established by addressing an over-identifying restriction problem with Generalized Method of Moments (GMM). Based on this finding, we have designed a new test statistic, and its asymptotic distribution has been derived. Empirical results across various datasets show that our method consistently outperforms existing ones.

Can the relative performance of a pre-trained large multimodal model (LMM) be predicted without access to labels? As LMMs proliferate, it becomes increasingly important to develop efficient ways to choose between them when faced with new data or tasks. The usual approach does the equivalent of giving the models an exam and marking them. We opt to avoid marking and the associated labor of determining the ground-truth answers. Instead, we explore other signals elicited and ascertain how well the models know their own limits, evaluating the effectiveness of these signals at unsupervised model ranking. We evaluate 47 state-of-the-art LMMs (e.g., LLaVA) across 9 visual question answering benchmarks, analyzing how well uncertainty-based metrics can predict relative model performance. Our findings show that uncertainty scores derived from softmax distributions provide a robust and consistent basis for ranking models across various tasks. This facilitates the ranking of LMMs on unlabeled data, providing a practical approach for selecting models for diverse target domains without requiring manual annotation.

LLM-agent systems often decompose high-level objectives into subtask dependency graphs, assuming that each subtask’s output is reliable and conditionally independent of others given its parent responses. However, this assumption frequently breaks during execution, as ground-truth responses are inaccessible, leading to inter-agent misalignment—failures caused by inconsistencies and coordination breakdowns among agents. To address this, we propose SeqCV, a dynamic framework for reliable execution under violated conditional independence. SeqCV executes subtasks sequentially, each conditioned on all prior verified responses, and performs consistency checks immediately after agents generate short token sequences. At each checkpoint, a token sequence is accepted only if it represents shared knowledge consistently supported across diverse LLM models; otherwise, it is discarded, triggering recursive subtask decomposition for finer-grained reasoning. Despite its sequential nature, SeqCV avoids repeated corrections on the same misalignment and achieves higher effective throughput than parallel pipelines. Across multiple reasoning and coordination tasks, SeqCV improves accuracy by up to 30\% over existing LLM-agent systems. Code is available at https://github.com/tmllab/2025_NeurIPS_SeqCV.

Text-to-Audio-Video (T2AV) generation aims to produce temporally and semantically aligned visual and auditory content from natural language descriptions. While recent progress in text-to-audio and text-to-video models has improved generation quality within each modality, jointly modeling them remains challenging due to incomplete and asymmetric correspondence: audio often reflects only a subset of the visual scene, and vice versa. Naively enforcing full alignment introduces semantic noise and temporal mismatches. To address this, we propose a novel framework that performs selective cross-modal alignment through a learnable masking mechanism, enabling the model to isolate and align only the shared latent components relevant to both modalities. This mechanism is integrated into an adaptation module that interfaces with pretrained encoders and decoders from latent video and audio diffusion models, preserving their generative capacity with reduced training overhead. Theoretically, we show that our masked objective provably recovers the minimal set of shared latent variables across modalities. Empirically, our method achieves state-of-the-art performance on standard T2AV benchmarks, demonstrating significant improvements in audiovisual synchronization and semantic consistency.

The transition matrix reveals the transition relationship between clean labels and noisy labels. It plays an important role in building statistically consistent classifiers for learning with noisy labels. However, in real-world applications, the transition matrix is usually unknown and has to be estimated. It is a challenging task to accurately estimate the transition matrix which usually depends on the instance. With both instances and noisy labels at hand, the major difficulty of estimating the transition matrix comes from the absence of clean label information. Recent work suggests that self-supervised learning methods can effectively infer clean label information. These methods could even achieve comparable performance with supervised learning on many benchmark datasets but without requiring any labels. Motivated by this, our paper presents a practical approach that harnesses self-supervised learning to extract clean label information, which reduces the estimation error of the instance-dependent transition matrix. By exploiting the estimated transition matrix, the performance of classifiers is improved. Empirical results on different datasets illustrate that our proposed methodology outperforms existing state-of-the-art methods in terms of both classification accuracy and transition matrix estimation.

The way research and business manage and utilize knowledge is undergoing a significant transformation, driven by Artificial Intelligence (AI). Deep learning and machine learning are emerging as powerful tools for optimizing knowledge management systems, leading to more informed and productive development. AI offers unique solutions for organizations struggling with information overload and inefficient knowledge transfer. These AI models can significantly improve data management and utilization. Imagine an AI-powered system that streamlines onboarding processes, provides precise answers to various queries, and even captures the valuable tacit knowledge (implicit skills and expertise) often residing within individuals. AI bridges the gap between explicit knowledge (easily documented information) and tacit knowledge, fostering a more comprehensive and accessible knowledge base. However, such AI systems solicit trustworthy and responsible approaches to mitigate potential misuse and malfunction. In this workshop, we aim to gather researchers and engineers from academia and industry to discuss the latest advances in trustworthy and responsible AI solutions for information and knowledge management systems.

Non-transferable learning (NTL) aims to restrict the generalization of models toward the target domain(s). To this end, existing works learn non-transferable representations by reducing statistical dependence between the source and target domain. However, such statistical methods essentially neglect to distinguish between *styles* and *contents*, leading them to inadvertently fit (i) spurious correlation between *styles* and *labels*, and (ii) fake independence between *contents* and *labels*. Consequently, their performance will be limited when natural distribution shifts occur or malicious intervention is imposed. In this paper, we propose a novel method (dubbed as H-NTL) to understand and advance the NTL problem by introducing a causal model to separately model *content* and *style* as two latent factors, based on which we disentangle and harness them as guidances for learning non-transferable representations with intrinsically causal relationships. Specifically, to avoid fitting spurious correlation and fake independence, we propose a variational inference framework to disentangle the naturally mixed *content factors* and *style factors* under our causal model. Subsequently, based on dual-path knowledge distillation, we harness the disentangled two *factors* as guidances for non-transferable representation learning: (i) we constraint the source domain representations to fit *content factors* (which are the intrinsic cause of *labels*), and (ii) we enforce that the target domain representations fit *style factors* which barely can predict labels. As a result, the learned feature representations follow optimal untransferability toward the target domain and minimal negative influence on the source domain, thus enabling better NTL performance. Empirically, the proposed H-NTL significantly outperforms competing methods by a large margin.

Contrastive learning, while highly effective for a lot of tasks, shows limited improvement in ordinal regression. We find that the limitation comes from the predefined strong data augmentations employed in contrastive learning. Intuitively, for ordinal regression datasets, the discriminative information (ordinal content information) contained in instances is subtle. The strong augmentations can easily overshadow or diminish this ordinal content information. As a result, when contrastive learning is used to extract common features between weakly and strongly augmented images, the derived features often lack this essential ordinal content, rendering them less useful in training models for ordinal regression. To improve contrastive learning's utility for ordinal regression, we propose a novel augmentation method to replace the predefined strong argumentation based on the principle of minimal change. Our method is designed in a generative manner that can effectively generate images with different styles but contains desired ordinal content information. Extensive experiments validate the effectiveness of our proposed method, which serves as a plug-and-play solution and consistently improves the performance of existing state-of-the-art methods in ordinal regression tasks.

In label-noise learning, the noise transition matrix reveals how an instance transitions from its clean label to its noisy label. Accurately estimating an instance's noise transition matrix is crucial for estimating its clean label. However, when only a noisy dataset is available, noise transition matrices can be estimated only for some "special" instances. To leverage these estimated transition matrices to help estimate the transition matrices of other instances, it is essential to explore relations between the matrices of these "special" instances and those of others. Existing studies typically build the relation by explicitly defining the similarity between the estimated noise transition matrices of "special" instances and those of other instances. However, these similarity-based assumptions are hard to validate and may not align with real-world data. If these assumptions fail, both noise transition matrices and clean labels cannot be accurately estimated. In this paper, we found that by learning the latent causal structure governing the generating process of noisy data, we can estimate noise transition matrices without the need for similarity-based assumptions. Unlike previous generative label-noise learning methods, we consider causal relations between latent causal variables and model them with a learnable graphical model. Utilizing only noisy data, our method can effectively learn the latent causal structure. Experimental results on various noisy datasets demonstrate that our method achieves state-of-the-art performance in estimating noise transition matrices, which leads to improved classification accuracy. The code is available at: https://github.com/tmllab/2024_NeurIPS_CSGN.

The application of reinforcement learning (RL) involving interactions with individuals has grown significantly in recent years. These interactions, influenced by factors such as personal preferences and physiological differences, causally influence state transitions, ranging from health conditions in healthcare to learning progress in education. As a result, different individuals may exhibit different state-transition processes. Understanding individualized state-transition processes is essential for optimizing individualized policies. In practice, however, identifying these state-transition processes is challenging, as individual-specific factors often remain latent. In this paper, we establish the identifiability of these latent factors and introduce a practical method that effectively learns these processes from observed state-action trajectories. Experiments on various datasets show that the proposed method can effectively identify latent state-transition processes and facilitate the learning of individualized RL policies.

Abstract Deep models trained by using clean data have achieved tremendous success in fine-grained image classification. Yet, they generally suffer from significant performance degradation when encountering noisy labels. Existing approaches to handle label noise, though proved to be effective for generic object recognition, usually fail on fine-grained data. The reason is that, on fine-grained data, the category difference is subtle and the training sample size is small. Then deep models could easily overfit the noisy labels. To improve the robustness of deep models on noisy data for fine-grained visual categorization, in this paper, we propose a novel learning framework named ProtoSimi. Our method employs an adaptive label correction strategy, ensuring effective learning on limited data. Specifically, our approach considers the criteria of exploring the effectiveness of both global class-prototype and part class-prototype similarities in identifying and correcting labels of samples. We evaluate our method on three standard benchmarks of fine-grained recognition. Experimental results show that our method outperforms the existing label noisy methods by a large margin. In ablation studies, we also verify that our method is non-sensitive to hyper-parameters selection and can be integrated with other FGVC methods to increase the generalization performance.

In real life, accurately annotating large-scale datasets is sometimes difficult. Datasets used for training deep learning models are likely to contain label noise. To make use of the dataset containing label noise, two typical methods have been proposed. One is to employ the semi-supervised method by exploiting labeled confident examples and unlabeled unconfident examples. The other one is to model label noise and design statistically consistent classifiers. A natural question remains unsolved: which one should be used for a specific real-world application? In this paper, we answer the question from the perspective of causal data generative process. Specifically, the performance of the semi-supervised based method depends heavily on the data generative process while the method modeling label-noise is not influenced by the generation process. For example, for a given dataset, if it has a causal generative structure that the features cause the label, the semi-supervised based method would not be helpful. When the causal structure is unknown, we provide an intuitive method to discover the causal structure for a given dataset containing label noise.

Relying on large-scale training data with pixel-level labels, previous edge detection methods have achieved high performance. However, it is hard to manually label edges accurately, especially for large datasets, and thus the datasets inevitably contain noisy labels. This label-noise issue has been studied extensively for classification, while still remaining under-explored for edge detection. To address the label-noise issue for edge detection, this paper proposes to learn Pixel-level Noise Transitions to model the label-corruption process. To achieve it, we develop a novel Pixel-wise Shift Learning (PSL) module to estimate the transition from clean to noisy labels as a displacement field. Exploiting the estimated noise transitions, our model, named PNT-Edge, is able to fit the prediction to clean labels. In addition, a local edge density regularization term is devised to exploit local structure information for better transition learning. This term encourages learning large shifts for the edges with complex local structures. Experiments on SBD and Cityscapes demonstrate the effectiveness of our method in relieving the impact of label noise. Codes will be available at github.com/DREAMXFAR/PNT-Edge.

Label noise widely exists in large-scale image datasets. To mitigate the side effects of label noise, state-of-the-art methods focus on selecting confident examples by leveraging semi-supervised learning. Existing research shows that the ability to extract hard confident examples, which are close to the decision boundary, significantly influences the generalization ability of the learned classifier. In this paper, we find that a key reason for some hard examples being close to the decision boundary is due to the entanglement of style factors with content factors. The hard examples become more discriminative when we focus solely on content factors, such as semantic information, while ignoring style factors. Nonetheless, given only noisy data, content factors are not directly observed and have to be inferred. To tackle the problem of inferring content factors for classification when learning with noisy labels, our objective is to ensure that the content factors of all examples in the same underlying clean class remain unchanged as their style information changes. To achieve this, we utilize different data augmentation techniques to alter the styles while regularizing content factors based on some confident examples. By training existing methods with our inferred content factors, CS-Isolate proves their effectiveness in learning hard examples on benchmark datasets. The implementation is available at https://github.com/tmllab/2023_NeurIPS_CS-isolate.

Deep learning algorithms have led to a series of breakthroughs in computer vision, acoustical signal processing, and others. However, they have only been popularized recently due to the groundbreaking techniques developed for training deep architectures. Understanding the training techniques is important if we want to further improve them. Through extensive experimentation, Erhan et al. (2010) empirically illustrated that unsupervised pretraining has an effect of regularization for deep learning algorithms. However, theoretical justifications for the observation remain elusive. In this article, we provide theoretical supports by analyzing how unsupervised pretraining regularizes deep learning algorithms. Specifically, we interpret deep learning algorithms as the traditional Tikhonov-regularized batch learning algorithms that simultaneously learn predictors in the input feature spaces and the parameters of the neural networks to produce the Tikhonov matrices. We prove that unsupervised pretraining helps in learning meaningful Tikhonov matrices, which will make the deep learning algorithms uniformly stable and the learned predictor will generalize fast w.r.t. the sample size. Unsupervised pretraining, therefore, can be interpreted as to have the function of regularization.

Label noise generally degenerates the performance of deep learning algorithms because deep neural networks easily overfit label errors. Let $X$ and $Y$ denote the instance and clean label, respectively. When $Y$ is a cause of $X$, according to which many datasets have been constructed, e.g., \textit{SVHN} and \textit{CIFAR}, the distributions of $P(X)$ and $P(Y|X)$ are generally entangled. This means that the unsupervised instances are helpful to learn the classifier and thus reduce the side effect of label noise. However, it remains elusive on how to exploit the causal information to handle the label-noise problem. We propose to model and make use of the causal process in order to correct the label-noise effect.Empirically, the proposed method outperforms all state-of-the-art methods on both synthetic and real-world label-noise datasets.

The transition matrix, denoting the transition relationship from clean labels to noisy labels, is essential to build statistically consistent classifiers in label-noise learning. Existing methods for estimating the transition matrix rely heavily on estimating the noisy class posterior. However, the estimation error for noisy class posterior could be large because of the randomness of label noise. The estimation error would lead the transition matrix to be poorly estimated. Therefore in this paper, we aim to solve this problem by exploiting the divide-and-conquer paradigm. Specifically, we introduce an intermediate class to avoid directly estimating the noisy class posterior. By this intermediate class, the original transition matrix can then be factorized into the product of two easy-to-estimated transition matrices. We term the proposed method as the dual $T$-estimator. Both theoretical analyses and empirical results illustrate the effectiveness of the dual $T$-estimator for estimating transition matrices, leading to better classification performances.