Large Multimodal Models (LMMs) face notable challenges when encountering multimodal knowledge conflicts, particularly under retrieval-augmented generation (RAG) frameworks, where the contextual information from external sources may contradict the model’s internal parametric knowledge, leading to unreliable outputs. However, existing benchmarks fail to reflect such realistic conflict scenarios. Most focus solely on intra-memory conflicts, while context-memory and inter-context conflicts remain largely unaddressed. Furthermore, commonly used factual knowledge-based evaluations are often overlooked, and existing datasets lack a thorough investigation into conflict detection capabilities.To bridge this gap, we propose MMKC-Bench, a benchmark designed to evaluate factual knowledge conflicts in both context-memory and inter-context scenarios. MMKC-Bench encompasses four types of multimodal knowledge conflicts and includes 1,881 knowledge instances and 3,997 images across 32 broad types, collected through automated pipelines with human verification. We evaluate four representative series of LMMs on both model behavior analysis and conflict detection tasks. Our findings show that while current LMMs are capable of recognizing knowledge conflicts, they tend to favor internal parametric knowledge over external evidence. We hope MMKC-Bench will foster further research in multimodal knowledge conflict and enhance the development of multimodal RAG systems.

Pre-trained Graph Neural Networks (GNNs) have demonstrated remarkable performance in graph mining tasks, yet they often amplify societal biases against protected demographic groups. Existing fairness-aware approaches primarily address discrimination based on a single sensitive attribute (e.g., gender or race), overlooking real-world scenarios where individuals possess multiple overlapping demographic characteristics, leading to unfair treatment of underrepresented subgroups. Moreover, incorporating extra fairness constraints into pre-trained GNNs usually requires full model retraining, which is computationally expensive and often impractical. To address these limitations, we propose a novel Multi-attribute Fairness-aware Prompt Tuning framework named MultiFPT. Our approach operates in two key stages: in the graph prompt learning stage, MultiFPT injects fairness-aware structural and feature prompts into pre-trained GNN inputs; in the adapter tuning stage, a lightweight adapter regularized by the Hilbert–Schmidt Independence Criterion (HSIC) enforces statistical independence between node representations and multiple sensitive attributes. Experiments on real-world datasets demonstrate that MultiFPT significantly improves multi-attribute fairness, reducing bias by approximately 30% on average in node classification while maintaining competitive predictive performance compared to state-of-the-art baselines.

Learning with noisy labels (LNL) methods have enabled the deployment of machine learning systems with imperfectly labeled data. However, these methods often struggle to identify noise in the presence of long-tailed (LT) class distributions, where the memorization effect becomes class-dependent. Conversely, LT methods are suboptimal under label noise, as it hinders access to accurate label frequency statistics. This study aims to address the long-tailed noisy data by bridging the methodological gap between LNL and LT approaches. We propose a direct solution, termed Robust Logit Adjustment, which estimates ground-truth labels through label refurbishment, thereby mitigating the impact of label noise. Simultaneously, our method incorporates the distribution of training-time corrected target labels into the LT method logit adjustment, providing class-rebalanced supervision. Extensive experiments on both synthetic and real-world long-tailed noisy datasets demonstrate the superior performance of our method.

Graph Neural Networks (GNNs) have demonstrated remarkable ability in semi-supervised node classification. However, most existing GNNs rely heavily on a large amount of labeled data for training, which is labor-intensive and requires extensive domain knowledge. In this paper, we first analyze the restrictions of GNNs generalization from the perspective of supervision signals in the context of few-shot semi-supervised node classification. To address these challenges, we propose a novel algorithm named NormProp, which utilizes the homophily assumption of unlabeled nodes to generate additional supervision signals, thereby enhancing the generalization against label scarcity. The key idea is to efficiently capture both the class information and the consistency of aggregation during message passing, via decoupling the direction and Euclidean norm of node representations. Moreover, we conduct a theoretical analysis to determine the upper bound of Euclidean norm, and then propose homophilous regularization to constraint the consistency of unlabeled nodes. Extensive experiments demonstrate that NormProp achieve state-of-the-art performance under low-label rate scenarios with low computational complexity.

Modern machine learning applications are characterized by the increasing size of deep models and the growing diversity of data modalities. This trend underscores the importance of efficiently adapting pre-trained multi-modal models to the test distribution in real time, i.e., multi-modal test-time adaptation. In practice, the magnitudes of multi-modal shifts vary because multiple data sources interact with the impact factor in diverse manners. In this research, we investigate the the under-explored practical scenario uni-modal distribution shift, where the distribution shift influences only one modality, leaving the others unchanged. Through theoretical and empirical analyses, we demonstrate that the presence of such shift impedes multi-modal fusion and leads to the negative transfer phenomenon in existing test-time adaptation techniques. To flexibly combat this unique shift, we propose a selective adaptation schema that incorporates multiple modality-specific adapters to accommodate potential shifts and a “router” module that determines which modality requires adaptation. Finally, we validate the effectiveness of our proposed method through extensive experimental evaluations. Code available at https://github.com/chenmc1996/Uni-Modal-Distribution-Shift.

Real-world machine learning applications seldom provide perfect labeled data, posing a challenge in developing models robust to noisy labels. Recent methods prioritize noise filtering based on the discrepancies between model predictions and the provided noisy labels, assuming samples with minimal classification losses to be clean. In this work, we capitalize on the consistency between the learned model and the complete noisy dataset, employing the data’s rich representational and topological information. We introduce LaplaceConfidence, a method that to obtain label confidence (i.e., clean probabilities) utilizing the Laplacian energy. Specifically, it first constructs graphs based on the feature representations of all noisy samples and minimizes the Laplacian energy to produce a low-energy graph. Clean labels should fit well into the low-energy graph while noisy ones should not, allowing our method to determine data’s clean probabilities. Furthermore, LaplaceConfidence is embedded into a holistic method for robust training, where co-training technique generates unbiased label confidence and label refurbishment technique better utilizes it. We also explore the dimensionality reduction technique to accommodate our method on large-scale noisy datasets. Our experiments demonstrate that LaplaceConfidence outperforms state-of-the-art methods on benchmark datasets under both synthetic and real-world noise. Code available at https://github.com/chenmc1996/LaplaceConfidence .

Learning with noisy labels (LNL) has attracted significant attention from the research community. Many recent LNL methods rely on the assumption that clean samples tend to have a "small loss." However, this assumption often fails to generalize to some real-world cases with imbalanced subpopulations, that is, training subpopulations that vary in sample size or recognition difficulty. Therefore, recent LNL methods face the risk of misclassifying those "informative" samples (e.g., hard samples or samples in the tail subpopulations) into noisy samples, leading to poor generalization performance. To address this issue, we propose a novel LNL method to deal with noisy labels and imbalanced subpopulations simultaneously. It first leverages sample correlation to estimate samples' clean probabilities for label correction and then utilizes corrected labels for distributionally robust optimization (DRO) to further improve the robustness. Specifically, in contrast to previous works using classification loss as the selection criterion, we introduce a feature-based metric that takes the sample correlation into account for estimating samples' clean probabilities. Then, we refurbish the noisy labels using the estimated clean probabilities and the pseudo-labels from the model's predictions. With refurbished labels, we use DRO to train the model to be robust to subpopulation imbalance. Extensive experiments on a wide range of benchmarks demonstrate that our technique can consistently improve state-of-the-art (SOTA) robust learning paradigms against noisy labels, especially when encountering imbalanced subpopulations. We provide our code in https://github.com/chenmc1996/LNL-IS.

Modern neural networks are known to give overconfident predictions for out-of-distribution inputs when deployed in the open world. It is common practice to leverage a surrogate outlier dataset to regularize the model during training, and recent studies emphasize the role of uncertainty in designing the sampling strategy for outlier datasets. However, the OOD samples selected solely based on predictive uncertainty can be biased towards certain types, which may fail to capture the full outlier distribution. In this work, we empirically show that diversity is critical in sampling outliers for OOD detection performance. Motivated by the observation, we propose a straightforward and novel sampling strategy named DOS (Diverse Outlier Sampling) to select diverse and informative outliers. Specifically, we cluster the normalized features at each iteration, and the most informative outlier from each cluster is selected for model training with absent category loss. With DOS, the sampled outliers efficiently shape a globally compact decision boundary between ID and OOD data. Extensive experiments demonstrate the superiority of DOS, reducing the average FPR95 by up to 25.79% on CIFAR-100 with TI-300K.

Parameter-efficient transfer learning (PETL) methods show promise in adapting a pre-trained model to various downstream tasks while training only a few parameters. In the computer vision (CV) domain, numerous PETL algorithms have been proposed, but their direct employment or comparison remains inconvenient. To address this challenge, we construct a Unified Visual PETL Benchmark (V-PETL Bench) for the CV domain by selecting 30 diverse, challenging, and comprehensive datasets from image recognition, video action recognition, and dense prediction tasks. On these datasets, we systematically evaluate 25 dominant PETL algorithms and open-source a modular and extensible codebase for fair evaluation of these algorithms. V-PETL Bench runs on NVIDIA A800 GPUs and requires approximately 310 GPU days. We release all the benchmark, making it more efficient and friendly to researchers. Additionally, V-PETL Bench will be continuously updated for new PETL algorithms and CV tasks.

Co-training, an advanced form of self-training, allows multiple base models to learn collaboratively, leading to superior performance in semi-supervised learning tasks. However, its widespread adoption is hindered by high computational costs and intricate design choices. To address these challenges, we present Multi-Head Co-Training, a streamlined and efficient framework that consolidates individual models into a multi-head structure, adding minimal extra parameters. Each classification head in this unified model collaborates with others via a “Weak and Strong Augmentation” strategy, with diversity organically introduced through robust data augmentation . Consequently, our approach implicitly promotes diversity while incurring only a minor increase in computational overhead, making co-training more accessible. We validate the effectiveness of Multi-Head Co-Training through an empirical study on standard semi-supervised learning benchmarks. For example, our method achieves up to a 3.1% accuracy improvement on the semi-supervised CIFAR dataset compared to recent methods. Recognizing the necessity for more practical performance metrics beyond accuracy, we assess our framework from three additional perspectives: robust generalization, uncertainty, and computational efficiency. To evaluate robust generalization, we expand the conventional SSL experimental setting to a more comprehensive open-set semi-supervised learning scenario. For uncertainty assessment, we conduct experiments on model calibration and selective classification benchmarks. For example, our method achieves up to a 4.3% accuracy improvement on the open-set semi-supervised CIFAR dataset. Our extensive experiments confirm that our proposed framework better captures prediction confidence and uncertainty, rendering it more suitable for SSL deployment in open environments. The code is available at https://github.com/chenmc1996/Multi-Head-Co-Training .

Noisy labels damage the performance of deep networks. For robust learning, a prominent two-stage pipeline alternates between eliminating possible incorrect labels and semi-supervised training. However, discarding part of noisy labels could result in a loss of information, especially when the corruption has a dependency on data, e.g., class-dependent or instance-dependent. Moreover, from the training dynamics of a representative two-stage method DivideMix, we identify the domination of confirmation bias: pseudo-labels fail to correct a considerable amount of noisy labels, and consequently, the errors accumulate. To sufficiently exploit information from noisy labels and mitigate wrong corrections, we propose Robust Label Refurbishment (Robust LR)—a new hybrid method that integrates pseudo-labeling and confidence estimation techniques to refurbish noisy labels. We show that our method successfully alleviates the damage of both label noise and confirmation bias. As a result, it achieves state-of-the-art performance across datasets and noise types, namely CIFAR under different levels of synthetic noise and mini-WebVision and ANIMAL-10N with real-world noise.

Detecting out-of-distribution (OOD) samples is crucial to the safe deployment of a classifier in the real world. However, deep neural networks are known to be overconfident for abnormal data. Existing works directly design score function by mining the inconsistency from classifier for in-distribution (ID) and OOD. In this paper, we further complement this inconsistency with reconstruction error, based on the assumption that an autoencoder trained on ID data cannot reconstruct OOD as well as ID. We propose a novel method, READ (Reconstruction Error Aggregated Detector), to unify inconsistencies from classifier and autoencoder. Specifically, the reconstruction error of raw pixels is transformed to latent space of classifier. We show that the transformed reconstruction error bridges the semantic gap and inherits detection performance from the original. Moreover, we propose an adjustment strategy to alleviate the overconfidence problem of autoencoder according to a fine-grained characterization of OOD data. Under two scenarios of pre-training and retraining, we respectively present two variants of our method, namely READ-MD (Mahalanobis Distance) only based on pre-trained classifier and READ-ED (Euclidean Distance) which retrains the classifier. Our methods do not require access to test time OOD data for fine-tuning hyperparameters. Finally, we demonstrate the effectiveness of the proposed methods through extensive comparisons with state-of-the-art OOD detection algorithms. On a CIFAR-10 pre-trained WideResNet, our method reduces the average FPR@95TPR by up to 9.8% compared with previous state-of-the-art.

Heterogeneous Information Networks (HINs) are prevalent in real-world systems. Recent advances in network embedding provide an effective way of encoding HINs into low-dimensional vectors. However, there is a growing concern that existing HIN embedding algorithms may suffer from the problem of generating biased representations, resulting in discrimination against certain demographic groups. In this paper, we propose a flexible debiasing framework for fair HIN embedding to address this issue. Specifically, we first formalize measurements and the definition of fairness in HIN embedding. Then, we propose a debiasing framework named FairHGNN, including a novel meta-path sampling method that focuses on mitigating the bias in random walks, and a fairness constraint with Wasserstein distance to alleviate the algorithmic bias in Graph Neural Networks (GNNs). Experimental results on real-world datasets validate the efficacy of FairHGNN in promoting fairness and maintaining good utility.

Immune repertoire classification, a typical multiple instance learning (MIL) problem, is a frontier research topic in computational biology that makes transformative contributions to new vaccines and immune therapies. However, the traditional instance-space MIL, directly assigning bag-level labels to instances, suffers from the massive amount of noisy labels and extremely low witness rate. In this work, we propose a noisy-label-learning formulation to solve the immune repertoire classification task. To remedy the inaccurate supervision of repertoire-level labels for a sequence-level classifier, we design a robust training strategy: The initial labels are smoothed to be asymmetric and are progressively corrected using the model's predictions throughout the training process. Furthermore, two models with the same architecture but different parameter initialization are co-trained simultaneously to remedy the known ``confirmation bias'' problem in the self-training-like schema. As a result, we obtain accurate sequence-level classification and, subsequently, repertoire-level classification. Experiments on the Cytomegalovirus (CMV) and Cancer datasets demonstrate our method's effectiveness and superior performance on sequence-level and repertoire-level tasks. Code available at https://github.com/TencentAILabHealthcare/NLL-IRC.

Out-of-distribution (OOD) detection is critical for ensuring the safe deployment of machine learning models in the open world. Due to the simplicity and intuitiveness of distance- based methods, i.e., samples are detected as OOD if they are relatively far away from the centroids or prototypes of in-distribution (ID) classes, they have attracted widespread attention from researchers in the field of OOD detection. However, prior OOD detection methods directly take off-the- shelf loss functions, like widely used softmax cross-entropy (CE) loss, that suffices for classifying ID samples, but is not optimally designed for OOD detection. In this work, we propose CESED, an improved CE loss applied to the scalable Squared Euclidean Distance vector, which exploits hyper- spherical evenly-distributed class centroids for OOD detection. CESED can promote strong ID-OOD separability because it explicitly encourages maximization of inter-class distances and minimization of intra-class distances. Extensive experiments demonstrate that CESED achieves superior detection performance on a comprehensive suite of benchmark datasets. For the more challenging case where CIFAR-100 is used as ID, our method achieves a 31.98% reduction in average FPR95 and 6.20% reduction in ID test error compared to the baseline method using a softmax confidence score.

Co-training, extended from self-training, is one of the frameworks for semi-supervised learning. Without natural split of features, single-view co-training works at the cost of training extra classifiers, where the algorithm should be delicately designed to prevent individual classifiers from collapsing into each other. To remove these obstacles which deter the adoption of single-view co-training, we present a simple and efficient algorithm Multi-Head Co-Training. By integrating base learners into a multi-head structure, the model is in a minimal amount of extra parameters. Every classification head in the unified model interacts with its peers through a “Weak and Strong Augmentation” strategy, in which the diversity is naturally brought by the strong data augmentation. Therefore, the proposed method facilitates single-view co-training by 1). promoting diversity implicitly and 2). only requiring a small extra computational overhead. The effectiveness of Multi-Head Co-Training is demonstrated in an empirical study on standard semi-supervised learning benchmarks.