Continual Learning in Large Language Models: Methods, Challenges, and Opportunities

Continual learning (CL) has emerged as a pivotal paradigm to enable large language models (LLMs) to dynamically adapt to evolving knowledge and sequential tasks while mitigating catastrophic forgetting-a critical limitation of the static pre-training paradigm inherent to modern LLMs. This survey presents a comprehensive overview of CL methodologies tailored for LLMs, structured around three core training stages: continual pre-training, continual fine-tuning, and continual alignment.Beyond the canonical taxonomy of rehearsal-, regularization-, and architecture-based methods, we further subdivide each category by its distinct forgetting mitigation mechanisms and conduct a rigorous comparative analysis of the adaptability and critical improvements of traditional CL methods for LLMs. In doing so, we explicitly highlight core distinctions between LLM CL and traditional machine learning, particularly with respect to scale, parameter efficiency, and emergent capabilities. Our analysis covers essential evaluation metrics, including forgetting rates and knowledge transfer efficiency, along with emerging benchmarks for assessing CL performance. This survey reveals that while current methods demonstrate promising results in specific domains, fundamental challenges persist in achieving seamless knowledge integration across diverse tasks and temporal scales. This systematic review contributes to the growing body of knowledge on LLM adaptation, providing researchers and practitioners with a structured framework for understanding current achievements and future opportunities in lifelong learning for language models.

C$^{2}$TC: A Training-Free Framework for Efficient Tabular Data Condensation

Tabular data is the primary data format in industrial relational databases, underpinning modern data analytics and decision-making. However, the increasing scale of tabular data poses significant computational and storage challenges to learning-based analytical systems. This highlights the need for data-efficient learning, which enables effective model training and generalization using substantially fewer samples. Dataset condensation (DC) has emerged as a promising data-centric paradigm that synthesizes small yet informative datasets to preserve data utility while reducing storage and training costs. However, existing DC methods are computationally intensive due to reliance on complex gradient-based optimization. Moreover, they often overlook key characteristics of tabular data, such as heterogeneous features and class imbalance. To address these limitations, we introduce C$^{2}$TC (Class-Adaptive Clustering for Tabular Condensation), the first training-free tabular dataset condensation framework that jointly optimizes class allocation and feature representation, enabling efficient and scalable condensation. Specifically, we reformulate the dataset condensation objective into a novel class-adaptive cluster allocation problem (CCAP), which eliminates costly training and integrates adaptive label allocation to handle class imbalance. To solve the NP-hard CCAP, we develop HFILS, a heuristic local search that alternates between soft allocation and class-wise clustering to efficiently obtain high-quality solutions. Moreover, a hybrid categorical feature encoding (HCFE) is proposed for semantics-preserving clustering of heterogeneous discrete attributes. Extensive experiments on 10 real-world datasets demonstrate that C$^{2}$TC improves efficiency by at least 2 orders of magnitude over state-of-the-art baselines, while achieving superior downstream performance.

HyperJoin: LLM-augmented Hypergraph Link Prediction for Joinable Table Discovery

As a pivotal task in data lake management, joinable table discovery has attracted widespread interest. While existing language model-based methods achieve remarkable performance by combining offline column representation learning with online ranking, their design insufficiently accounts for the underlying structural interactions: (1) offline, they directly model tables into isolated or pairwise columns, thereby struggling to capture the rich inter-table and intra-table structural information; and (2) online, they rank candidate columns based solely on query-candidate similarity, ignoring the mutual interactions among the candidates, leading to incoherent result sets. To address these limitations, we propose HyperJoin, a large language model (LLM)-augmented Hypergraph framework for Joinable table discovery. Specifically, we first construct a hypergraph to model tables using both the intra-table hyperedges and the LLM-augmented inter-table hyperedges. Consequently, the task of joinable table discovery is formulated as link prediction on this constructed hypergraph. We then design HIN, a Hierarchical Interaction Network that learns expressive column representations through bidirectional message passing over columns and hyperedges. To strengthen coherence and internal consistency in the result columns, we cast online ranking as a coherence-aware top-k column selection problem. We then introduce a reranking module that leverages a maximum spanning tree algorithm to prune noisy connections and maximize coherence. Experiments demonstrate the superiority of HyperJoin, achieving average improvements of 21.4% (Precision@15) and 17.2% (Recall@15) over the best baseline.

FADTI: Fourier and Attention Driven Diffusion for Multivariate Time Series Imputation

Multivariate time series imputation is fundamental in applications such as healthcare, traffic forecasting, and biological modeling, where sensor failures and irregular sampling lead to pervasive missing values. However, existing Transformer- and diffusion-based models lack explicit inductive biases and frequency awareness, limiting their generalization under structured missing patterns and distribution shifts. We propose FADTI, a diffusion-based framework that injects frequency-informed feature modulation via a learnable Fourier Bias Projection (FBP) module and combines it with temporal modeling through self-attention and gated convolution. FBP supports multiple spectral bases, enabling adaptive encoding of both stationary and non-stationary patterns. This design injects frequency-domain inductive bias into the generative imputation process. Experiments on multiple benchmarks, including a newly introduced biological time series dataset, show that FADTI consistently outperforms state-of-the-art methods, particularly under high missing rates. Code is available at https://anonymous.4open.science/r/TimeSeriesImputation-52BF

DHG-Bench: A Comprehensive Benchmark on Deep Hypergraph Learning

Although conventional deep graph models have achieved great success in relational learning, their focus on pairwise relationships limits their capacity to learn pervasive higher-order interactions in real-world complex systems, which can be naturally modeled as hypergraphs. To tackle this, hypergraph neural networks (HNNs), the dominant approach in deep hypergraph learning (DHGL), has garnered substantial attention in recent years. Despite the proposal of numerous HNN methods, there is no comprehensive benchmark for HNNs, which creates a great obstacle to understanding the progress of DHGL in several aspects: (i) insufficient coverage of datasets, algorithms, and tasks; (ii) a narrow evaluation of algorithm performance; and (iii) inconsistent dataset usage, preprocessing, and experimental setups that hinder comparability. To fill the gap, we introduce DHG-Bench, the first comprehensive benchmark for DHGL. Specifically, DHG-Bench integrates 20 diverse datasets spanning node-, edge-, and graph-level tasks, along with 16 state-of-the-art HNN algorithms, under consistent data processing and experimental protocols. Our benchmark systematically investigates the characteristics of HNNs in terms of four dimensions: effectiveness, efficiency, robustness, and fairness. Further, to facilitate reproducible research, we have developed an easy-to-use library for training and evaluating different HNN methods. Extensive experiments conducted with DHG-Bench reveal both the strengths and inherent limitations of existing algorithms, offering valuable insights and directions for future research. The code is publicly available at: https://github.com/Coco-Hut/DHG-Bench.

OSS-UAgent: An Agent-based Usability Evaluation Framework for Open Source Software

Usability evaluation is critical to the impact and adoption of open source software (OSS), yet traditional methods relying on human evaluators suffer from high costs and limited scalability. To address these limitations, we introduce OSS-UAgent, an automated, configurable, and interactive agent-based usability evaluation framework specifically designed for open source software. Our framework employs intelligent agents powered by large language models (LLMs) to simulate developers performing programming tasks across various experience levels (from Junior to Expert). By dynamically constructing platform-specific knowledge bases, OSS-UAgent ensures accurate and context-aware code generation. The generated code is automatically evaluated across multiple dimensions, including compliance, correctness, and readability, providing a comprehensive measure of the software's usability. Additionally, our demonstration showcases OSS-UAgent's practical application in evaluating graph analytics platforms, highlighting its effectiveness in automating usability evaluation.

ProbDiffFlow: An Efficient Learning-Free Framework for Probabilistic Single-Image Optical Flow Estimation

This paper studies optical flow estimation, a critical task in motion analysis with applications in autonomous navigation, action recognition, and film production. Traditional optical flow methods require consecutive frames, which are often unavailable due to limitations in data acquisition or real-world scene disruptions. Thus, single-frame optical flow estimation is emerging in the literature. However, existing single-frame approaches suffer from two major limitations: (1) they rely on labeled training data, making them task-specific, and (2) they produce deterministic predictions, failing to capture motion uncertainty. To overcome these challenges, we propose ProbDiffFlow, a training-free framework that estimates optical flow distributions from a single image. Instead of directly predicting motion, ProbDiffFlow follows an estimation-by-synthesis paradigm: it first generates diverse plausible future frames using a diffusion-based model, then estimates motion from these synthesized samples using a pre-trained optical flow model, and finally aggregates the results into a probabilistic flow distribution. This design eliminates the need for task-specific training while capturing multiple plausible motions. Experiments on both synthetic and real-world datasets demonstrate that ProbDiffFlow achieves superior accuracy, diversity, and efficiency, outperforming existing single-image and two-frame baselines.

Unlocking Multi-Modal Potentials for Dynamic Text-Attributed Graph Representation

Dynamic Text-Attributed Graphs (DyTAGs) are a novel graph paradigm that captures evolving temporal edges alongside rich textual attributes. A prior approach to representing DyTAGs leverages pre-trained language models to encode text attributes and subsequently integrates them into dynamic graph models. However, it follows edge-centric modeling, as in dynamic graph learning, which is limited in local structures and fails to exploit the unique characteristics of DyTAGs, leading to suboptimal performance. We observe that DyTAGs inherently comprise three distinct modalities-temporal, textual, and structural-often exhibiting dispersed or even orthogonal distributions, with the first two largely overlooked in existing research. Building on this insight, we propose MoMent, a model-agnostic multi-modal framework that can seamlessly integrate with dynamic graph models for structural modality learning. The core idea is to shift from edge-centric to node-centric modeling, fully leveraging three modalities for node representation. Specifically, MoMent presents non-shared node-centric encoders based on the attention mechanism to capture global temporal and semantic contexts from temporal and textual modalities, together with local structure learning, thus generating modality-specific tokens. To prevent disjoint latent space, we propose a symmetric alignment loss, an auxiliary objective that aligns temporal and textual tokens, ensuring global temporal-semantic consistency with a theoretical guarantee. Last, we design a lightweight adaptor to fuse these tokens, generating comprehensive and cohesive node representations. We theoretically demonstrate that MoMent enhances discriminative power over exclusive edge-centric modeling. Extensive experiments across seven datasets and two downstream tasks show that MoMent achieves up to 33.62% improvement against the baseline using four dynamic graph models.

UniDyG: A Unified and Effective Representation Learning Approach for Large Dynamic Graphs

Dynamic graphs are formulated in continuous-time or discrete-time dynamic graphs. They differ in temporal granularity: Continuous-Time Dynamic Graphs (CTDGs) exhibit rapid, localized changes, while Discrete-Time Dynamic Graphs (DTDGs) show gradual, global updates. This difference leads to isolated developments in representation learning for each type. To advance representation learning, recent research attempts to design a unified model capable of handling both CTDGs and DTDGs. However, it typically focuses on local dynamic propagation for temporal structure learning in the time domain, failing to accurately capture the structural evolution associated with each temporal granularity. In addition, existing works-whether specific or unified-often overlook the issue of temporal noise, compromising the model robustness and effectiveness. To better model both types of dynamic graphs, we propose UniDyG, a unified and effective representation learning approach, which scales to large dynamic graphs. We first propose a novel Fourier Graph Attention (FGAT) mechanism that can model local and global structural correlations based on recent neighbors and complex-number selective aggregation, while theoretically ensuring consistent representations of dynamic graphs over time. Based on approximation theory, we demonstrate that FGAT is well-suited to capture the underlying structures in CTDGs and DTDGs. We further enhance FGAT to resist temporal noise by designing an energy-gated unit, which adaptively filters out high-frequency noise according to the energy. Last, we leverage our FGAT mechanisms for temporal structure learning and employ the frequency-enhanced linear function for node-level dynamic updates, facilitating the generation of high-quality temporal embeddings. Extensive experiments show that our UniDyG achieves an average improvement of 14.4% over sixteen baselines across nine dynamic graphs.

On LLM-Enhanced Mixed-Type Data Imputation with High-Order Message Passing

Missing data imputation, which aims to impute the missing values in the raw datasets to achieve the completeness of datasets, is crucial for modern data-driven models like large language models (LLMs) and has attracted increasing interest over the past decades. Despite its importance, existing solutions for missing data imputation either 1) only support numerical and categorical data or 2) show an unsatisfactory performance due to their design prioritizing text data and the lack of key properties for tabular data imputation. In this paper, we propose UnIMP, a Unified IMPutation framework that leverages LLM and high-order message passing to enhance the imputation of mixed-type data including numerical, categorical, and text data. Specifically, we first introduce a cell-oriented hypergraph to model the table. We then propose BiHMP, an efficient Bidirectional High-order Message-Passing network to aggregate global-local information and high-order relationships on the constructed hypergraph while capturing the inter-column heterogeneity and intra-column homogeneity. To effectively and efficiently align the capacity of the LLM with the information aggregated by BiHMP, we introduce Xfusion, which, together with BiHMP, acts as adapters for the LLM. We follow a pre-training and fine-tuning pipeline to train UnIMP, integrating two optimizations: chunking technique, which divides tables into smaller chunks to enhance efficiency; and progressive masking technique, which gradually adapts the model to learn more complex data patterns. Both theoretical proofs and empirical experiments on 10 real world datasets highlight the superiority of UnIMP over existing techniques.