arXiv subjects

Fan Lyu

Publications and source records attributed to Fan Lyu.

At least 19 recordsLinked to original sources

arXiv2026

Group Preference Collapse in Personalized Multimodal Large Language Models

Personalized multimodal large language models (MLLMs) aim to generate user-specific responses, but existing methods mainly rely on profile-level information and overlook diverse user preferences. We identify group preference collapse, where multi-user personalized MLLMs become insensitive to individual preferences and drift toward dominant population-level choices due to suppressed preference signals and unreliable preference use during generation. We propose PrefMoE, a preference-centric framework that separates stable profile information from preference-related representations. PrefMoE decomposes preferences into shared prototypes and personalized residuals, preserves individualized residuals with imbalance-aware learning, counterfactual pseudo-user augmentation, and residual decorrelation, and routes profile and preference factors through separate LoRA adaptation paths. Experiments across multiple MLLM backbones show that PrefMoE improves preference-sensitive personalization while substantially reducing preference collapse. Project page: https://prefmoe.github.io/.

cs.AI↗

arXiv2025

Annotation-Efficient Active Test-Time Adaptation with Conformal Prediction

Active Test-Time Adaptation (ATTA) improves model robustness under domain shift by selectively querying human annotations at deployment, but existing methods use heuristic uncertainty measures and suffer from low data selection efficiency, wasting human annotation budget. We propose Conformal Prediction Active TTA (CPATTA), which first brings principled, coverage-guaranteed uncertainty into ATTA. CPATTA employs smoothed conformal scores with a top-K certainty measure, an online weight-update algorithm driven by pseudo coverage, a domain-shift detector that adapts human supervision, and a staged update scheme balances human-labeled and model-labeled data. Extensive experiments demonstrate that CPATTA consistently outperforms the state-of-the-art ATTA methods by around 5% in accuracy. Our code and datasets are available at https://github.com/tingyushi/CPATTA.

cs.LG↗

arXiv2025

DeCLIP: Decoupled Prompting for Multi-Label Class-Incremental Learning with CLIP

Multi-label class-incremental learning (MLCIL) continuously expands the label space while recognizing multiple co-occurring categories, making catastrophic forgetting a central challenge. Recent class-incremental learning methods have increasingly adopted CLIP as their backbone. However, we find that applying CLIP to MLCIL exhibits two critical issues: entanglement of class-specific cues in shared visual representations and high false-positive rates (FPR) under task-level partial labeling. We propose DeCLIP, a replay-free and parameter-efficient framework for CLIP-based MLCIL. DeCLIP uses Decoupled Prompting to learn class-specific positive and negative prompts in both visual and textual modalities, enabling class-conditioned vision-language matching and reducing representation entanglement. Only new-category prompts are optimized, previous prompts remain unchanged, preserving prior knowledge and mitigating catastrophic forgetting without replay. DeCLIP further incorporates Adaptive Similarity Tempering, an inference-time strategy that adapts similarity-tempering strength to the incremental configuration, suppressing false positives without specific tuning. Experiments on MS-COCO, PASCAL VOC, and the real-world NUS-WIDEseq benchmark demonstrate consistent improvements over prior methods with a few trainable parameters.

cs.CV↗

arXiv2025

Ambiguity-Guided Learnable Distribution Calibration for Semi-Supervised Few-Shot Class-Incremental Learning

Few-Shot Class-Incremental Learning (FSCIL) focuses on models learning new concepts from limited data while retaining knowledge of previous classes. Recently, many studies have started to leverage unlabeled samples to assist models in learning from few-shot samples, giving rise to the field of Semi-supervised Few-shot Class-Incremental Learning (Semi-FSCIL). However, these studies often assume that the source of unlabeled data is only confined to novel classes of the current session, which presents a narrow perspective and cannot align well with practical scenarios. To better reflect real-world scenarios, we redefine Semi-FSCIL as Generalized Semi-FSCIL (GSemi-FSCIL) by incorporating both base and all the ever-seen novel classes in the unlabeled set. This change in the composition of unlabeled samples poses a new challenge for existing methods, as they struggle to distinguish between unlabeled samples from base and novel classes. To address this issue, we propose an Ambiguity-guided Learnable Distribution Calibration (ALDC) strategy. ALDC dynamically uses abundant base samples to correct biased feature distributions for few-shot novel classes. Experiments on three benchmark datasets show that our method outperforms existing works, setting new state-of-the-art results.

cs.CV↗

arXiv2025

MM-Prompt: Cross-Modal Prompt Tuning for Continual Visual Question Answering

Continual Visual Question Answering (CVQA) based on pre-trained models(PTMs) has achieved promising progress by leveraging prompt tuning to enable continual multi-modal learning. However, most existing methods adopt cross-modal prompt isolation, constructing visual and textual prompts separately, which exacerbates modality imbalance and leads to degraded performance over time. To tackle this issue, we propose MM-Prompt, a novel framework incorporating cross-modal prompt query and cross-modal prompt recovery. The former enables balanced prompt selection by incorporating cross-modal signals during query formation, while the latter promotes joint prompt reconstruction through iterative cross-modal interactions, guided by an alignment loss to prevent representational drift. Extensive experiments show that MM-Prompt surpasses prior approaches in accuracy and knowledge retention, while maintaining balanced modality engagement throughout continual learning.

cs.CV↗

arXiv2025

Sparse Tuning Enhances Plasticity in PTM-based Continual Learning

Continual Learning with Pre-trained Models holds great promise for efficient adaptation across sequential tasks. However, most existing approaches freeze PTMs and rely on auxiliary modules like prompts or adapters, limiting model plasticity and leading to suboptimal generalization when facing significant distribution shifts. While full fine-tuning can improve adaptability, it risks disrupting crucial pre-trained knowledge. In this paper, we propose Mutual Information-guided Sparse Tuning (MIST), a plug-and-play method that selectively updates a small subset of PTM parameters, less than 5%, based on sensitivity to mutual information objectives. MIST enables effective task-specific adaptation while preserving generalization. To further reduce interference, we introduce strong sparsity regularization by randomly dropping gradients during tuning, resulting in fewer than 0.5% of parameters being updated per step. Applied before standard freeze-based methods, MIST consistently boosts performance across diverse continual learning benchmarks. Experiments show that integrating our method into multiple baselines yields significant performance gains. Our code is available at https://github.com/zhwhu/MIST.

cs.LG↗

arXiv2025

Beyond Background Shift: Rethinking Instance Replay in Continual Semantic Segmentation

In this work, we focus on continual semantic segmentation (CSS), where segmentation networks are required to continuously learn new classes without erasing knowledge of previously learned ones. Although storing images of old classes and directly incorporating them into the training of new models has proven effective in mitigating catastrophic forgetting in classification tasks, this strategy presents notable limitations in CSS. Specifically, the stored and new images with partial category annotations leads to confusion between unannotated categories and the background, complicating model fitting. To tackle this issue, this paper proposes a novel Enhanced Instance Replay (EIR) method, which not only preserves knowledge of old classes while simultaneously eliminating background confusion by instance storage of old classes, but also mitigates background shifts in the new images by integrating stored instances with new images. By effectively resolving background shifts in both stored and new images, EIR alleviates catastrophic forgetting in the CSS task, thereby enhancing the model's capacity for CSS. Experimental results validate the efficacy of our approach, which significantly outperforms state-of-the-art CSS methods.

cs.CV↗

arXiv2025

Test-Time Discovery via Hashing Memory

We introduce Test-Time Discovery (TTD) as a novel task that addresses class shifts during testing, requiring models to simultaneously identify emerging categories while preserving previously learned ones. A key challenge in TTD is distinguishing newly discovered classes from those already identified. To address this, we propose a training-free, hash-based memory mechanism that enhances class discovery through fine-grained comparisons with past test samples. Leveraging the characteristics of unknown classes, our approach introduces hash representation based on feature scale and directions, utilizing Locality-Sensitive Hashing (LSH) for efficient grouping of similar samples. This enables test samples to be easily and quickly compared with relevant past instances. Furthermore, we design a collaborative classification strategy, combining a prototype classifier for known classes with an LSH-based classifier for novel ones. To enhance reliability, we incorporate a self-correction mechanism that refines memory labels through hash-based neighbor retrieval, ensuring more stable and accurate class assignments. Experimental results demonstrate that our method achieves good discovery of novel categories while maintaining performance on known classes, establishing a new paradigm in model testing. Our code is available at https://github.com/fanlyu/ttd.

cs.CV↗

arXiv2025

Conformal Uncertainty Indicator for Continual Test-Time Adaptation

Continual Test-Time Adaptation (CTTA) aims to adapt models to sequentially changing domains during testing, relying on pseudo-labels for self-adaptation. However, incorrect pseudo-labels can accumulate, leading to performance degradation. To address this, we propose a Conformal Uncertainty Indicator (CUI) for CTTA, leveraging Conformal Prediction (CP) to generate prediction sets that include the true label with a specified coverage probability. Since domain shifts can lower the coverage than expected, making CP unreliable, we dynamically compensate for the coverage by measuring both domain and data differences. Reliable pseudo-labels from CP are then selectively utilized to enhance adaptation. Experiments confirm that CUI effectively estimates uncertainty and improves adaptation performance across various existing CTTA methods.

cs.LG↗

arXiv2024

CALA: A Class-Aware Logit Adapter for Few-Shot Class-Incremental Learning

Few-Shot Class-Incremental Learning (FSCIL) defines a practical but challenging task where models are required to continuously learn novel concepts with only a few training samples. Due to data scarcity, existing FSCIL methods resort to training a backbone with abundant base data and then keeping it frozen afterward. However, the above operation often causes the backbone to overfit to base classes while overlooking the novel ones, leading to severe confusion between them. To address this issue, we propose Class-Aware Logit Adapter (CALA). Our method involves a lightweight adapter that learns to rectify biased predictions through a pseudo-incremental learning paradigm. In the real FSCIL process, we use the learned adapter to dynamically generate robust balancing factors. These factors can adjust confused novel instances back to their true label space based on their similarity to base classes. Specifically, when confusion is more likely to occur in novel instances that closely resemble base classes, greater rectification is required. Notably, CALA operates on the classifier level, preserving the original feature space, thus it can be flexibly plugged into most of the existing FSCIL works for improved performance. Experiments on three benchmark datasets consistently validate the effectiveness and flexibility of CALA. Codes will be available upon acceptance.

cs.CV↗

arXiv2024

Rebalancing Multi-Label Class-Incremental Learning

Multi-label class-incremental learning (MLCIL) is essential for real-world multi-label applications, allowing models to learn new labels while retaining previously learned knowledge continuously. However, recent MLCIL approaches can only achieve suboptimal performance due to the oversight of the positive-negative imbalance problem, which manifests at both the label and loss levels because of the task-level partial label issue. The imbalance at the label level arises from the substantial absence of negative labels, while the imbalance at the loss level stems from the asymmetric contributions of the positive and negative loss parts to the optimization. To address the issue above, we propose a Rebalance framework for both the Loss and Label levels (RebLL), which integrates two key modules: asymmetric knowledge distillation (AKD) and online relabeling (OR). AKD is proposed to rebalance at the loss level by emphasizing the negative label learning in classification loss and down-weighting the contribution of overconfident predictions in distillation loss. OR is designed for label rebalance, which restores the original class distribution in memory by online relabeling the missing classes. Our comprehensive experiments on the PASCAL VOC and MS-COCO datasets demonstrate that this rebalancing strategy significantly improves performance, achieving new state-of-the-art results even with a vanilla CNN backbone.

cs.CV↗

arXiv2024

Constructing Enhanced Mutual Information for Online Class-Incremental Learning

Online Class-Incremental continual Learning (OCIL) addresses the challenge of continuously learning from a single-channel data stream, adapting to new tasks while mitigating catastrophic forgetting. Recently, Mutual Information (MI)-based methods have shown promising performance in OCIL. However, existing MI-based methods treat various knowledge components in isolation, ignoring the knowledge confusion across tasks. This narrow focus on simple MI knowledge alignment may lead to old tasks being easily forgotten with the introduction of new tasks, risking the loss of common parts between past and present knowledge.To address this, we analyze the MI relationships from the perspectives of diversity, representativeness, and separability, and propose an Enhanced Mutual Information (EMI) method based on knwoledge decoupling. EMI consists of Diversity Mutual Information (DMI), Representativeness Mutual Information (RMI) and Separability Mutual Information (SMI). DMI diversifies intra-class sample features by considering the similarity relationships among inter-class sample features to enable the network to learn more general knowledge. RMI summarizes representative features for each category and aligns sample features with these representative features, making the intra-class sample distribution more compact. SMI establishes MI relationships for inter-class representative features, enhancing the stability of representative features while increasing the distinction between inter-class representative features, thus creating clear boundaries between class. Extensive experimental results on widely used benchmark datasets demonstrate the superior performance of EMI over state-of-the-art baseline methods.

cs.LG↗

arXiv2024

Towards stable training of parallel continual learning

Parallel Continual Learning (PCL) tasks investigate the training methods for continual learning with multi-source input, where data from different tasks are learned as they arrive. PCL offers high training efficiency and is well-suited for complex multi-source data systems, such as autonomous vehicles equipped with multiple sensors. However, at any time, multiple tasks need to be trained simultaneously, leading to severe training instability in PCL. This instability manifests during both forward and backward propagation, where features are entangled and gradients are conflict. This paper introduces Stable Parallel Continual Learning (SPCL), a novel approach that enhances the training stability of PCL for both forward and backward propagation. For the forward propagation, we apply Doubly-block Toeplit (DBT) Matrix based orthogonality constraints to network parameters to ensure stable and consistent propagation. For the backward propagation, we employ orthogonal decomposition for gradient management stabilizes backpropagation and mitigates gradient conflicts across tasks. By optimizing gradients by ensuring orthogonality and minimizing the condition number, SPCL effectively stabilizing the gradient descent in complex optimization tasks. Experimental results demonstrate that SPCL outperforms state-of-the-art methjods and achieve better training stability.

cs.LG↗

arXiv2024

Parameter-Selective Continual Test-Time Adaptation

Continual Test-Time Adaptation (CTTA) aims to adapt a pretrained model to ever-changing environments during the test time under continuous domain shifts. Most existing CTTA approaches are based on the Mean Teacher (MT) structure, which contains a student and a teacher model, where the student is updated using the pseudo-labels from the teacher model, and the teacher is then updated by exponential moving average strategy. However, these methods update the MT model indiscriminately on all parameters of the model. That is, some critical parameters involving sharing knowledge across different domains may be erased, intensifying error accumulation and catastrophic forgetting. In this paper, we introduce Parameter-Selective Mean Teacher (PSMT) method, which is capable of effectively updating the critical parameters within the MT network under domain shifts. First, we introduce a selective distillation mechanism in the student model, which utilizes past knowledge to regularize novel knowledge, thereby mitigating the impact of error accumulation. Second, to avoid catastrophic forgetting, in the teacher model, we create a mask through Fisher information to selectively update parameters via exponential moving average, with preservation measures applied to crucial parameters. Extensive experimental results verify that PSMT outperforms state-of-the-art methods across multiple benchmark datasets. Our code is available at \url{https://github.com/JiaxuTian/PSMT}.

cs.LG↗

arXiv2024

Less is More: Pseudo-Label Filtering for Continual Test-Time Adaptation

Continual Test-Time Adaptation (CTTA) aims to adapt a pre-trained model to a sequence of target domains during the test phase without accessing the source data. To adapt to unlabeled data from unknown domains, existing methods rely on constructing pseudo-labels for all samples and updating the model through self-training. However, these pseudo-labels often involve noise, leading to insufficient adaptation. To improve the quality of pseudo-labels, we propose a pseudo-label selection method for CTTA, called Pseudo Labeling Filter (PLF). The key idea of PLF is to keep selecting appropriate thresholds for pseudo-labels and identify reliable ones for self-training. Specifically, we present three principles for setting thresholds during continuous domain learning, including initialization, growth and diversity. Based on these principles, we design Self-Adaptive Thresholding to filter pseudo-labels. Additionally, we introduce a Class Prior Alignment (CPA) method to encourage the model to make diverse predictions for unknown domain samples. Through extensive experiments, PLF outperforms current state-of-the-art methods, proving its effectiveness in CTTA.

cs.LG↗

arXiv2024

Improving Data-aware and Parameter-aware Robustness for Continual Learning

The goal of Continual Learning (CL) task is to continuously learn multiple new tasks sequentially while achieving a balance between the plasticity and stability of new and old knowledge. This paper analyzes that this insufficiency arises from the ineffective handling of outliers, leading to abnormal gradients and unexpected model updates. To address this issue, we enhance the data-aware and parameter-aware robustness of CL, proposing a Robust Continual Learning (RCL) method. From the data perspective, we develop a contrastive loss based on the concepts of uniformity and alignment, forming a feature distribution that is more applicable to outliers. From the parameter perspective, we present a forward strategy for worst-case perturbation and apply robust gradient projection to the parameters. The experimental results on three benchmarks show that the proposed method effectively maintains robustness and achieves new state-of-the-art (SOTA) results. The code is available at: https://github.com/HanxiXiao/RCL

cs.LG↗

arXiv2024

Controllable Continual Test-Time Adaptation

Continual Test-Time Adaptation (CTTA) is an emerging and challenging task where a model trained in a source domain must adapt to continuously changing conditions during testing, without access to the original source data. CTTA is prone to error accumulation due to uncontrollable domain shifts, leading to blurred decision boundaries between categories. Existing CTTA methods primarily focus on suppressing domain shifts, which proves inadequate during the unsupervised test phase. In contrast, we introduce a novel approach that guides rather than suppresses these shifts. Specifically, we propose $\textbf{C}$ontrollable $\textbf{Co}$ntinual $\textbf{T}$est-$\textbf{T}$ime $\textbf{A}$daptation (C-CoTTA), which explicitly prevents any single category from encroaching on others, thereby mitigating the mutual influence between categories caused by uncontrollable shifts. Moreover, our method reduces the sensitivity of model to domain transformations, thereby minimizing the magnitude of category shifts. Extensive quantitative experiments demonstrate the effectiveness of our method, while qualitative analyses, such as t-SNE plots, confirm the theoretical validity of our approach.

cs.LG↗

arXiv2024

Overcoming Domain Drift in Online Continual Learning

Online Continual Learning (OCL) empowers machine learning models to acquire new knowledge online across a sequence of tasks. However, OCL faces a significant challenge: catastrophic forgetting, wherein the model learned in previous tasks is substantially overwritten upon encountering new tasks, leading to a biased forgetting of prior knowledge. Moreover, the continual doman drift in sequential learning tasks may entail the gradual displacement of the decision boundaries in the learned feature space, rendering the learned knowledge susceptible to forgetting. To address the above problem, in this paper, we propose a novel rehearsal strategy, termed Drift-Reducing Rehearsal (DRR), to anchor the domain of old tasks and reduce the negative transfer effects. First, we propose to select memory for more representative samples guided by constructed centroids in a data stream. Then, to keep the model from domain chaos in drifting, a two-level angular cross-task Contrastive Margin Loss (CML) is proposed, to encourage the intra-class and intra-task compactness, and increase the inter-class and inter-task discrepancy. Finally, to further suppress the continual domain drift, we present an optional Centorid Distillation Loss (CDL) on the rehearsal memory to anchor the knowledge in feature space for each previous old task. Extensive experimental results on four benchmark datasets validate that the proposed DRR can effectively mitigate the continual domain drift and achieve the state-of-the-art (SOTA) performance in OCL.

cs.LG↗