Multimodal large language models (MLLMs) have garnered widespread attention from researchers due to their remarkable understanding and generation capabilities in visual language tasks (e.g., visual question answering). However, the rapid pace of knowledge updates in the real world makes offline training of MLLMs costly, and when faced with non-stationary data streams, MLLMs suffer from catastrophic forgetting during learning. In this paper, we propose an MLLMs-based dual momentum Mixture-of-Experts (CL-MoE) framework for continual visual question answering (VQA). We integrate MLLMs with continual learning to utilize the rich commonsense knowledge in LLMs. We introduce a Dual-Router MoE (RMoE) strategy to select the global and local experts using task-level and instance-level routers, to robustly assign weights to the experts most appropriate for the task. Then, we design a dynamic Momentum MoE (MMoE) to update the parameters of experts dynamically based on the relationships between the experts and tasks/instances, so that the model can absorb new knowledge while maintaining existing knowledge. The extensive experimental results indicate that our method achieves state-of-the-art performance on 10 VQA tasks, proving the effectiveness of our approach.
Continual learning has become a promising solution to refine large language models incrementally by leveraging user feedback. In particular, online continual learning - iteratively training the model with small batches of user feedback - has demonstrated notable performance improvements. However, the existing practice of separating training and serving processes forces the online trainer to recompute the intermediate results already done during serving. Such redundant computations can account for 30%-42% of total training time. In this paper, we propose Alchemist, to the best of our knowledge, the first online continual learning system that efficiently reuses serving activations to increase training throughput. Alchemist introduces two key techniques: (1) recording and storing activations and KV cache only during the prefill phase to minimize latency and memory overhead; and (2) smart activation offloading and hedging. Evaluations with inputs of varied token length sampled from ShareGPT dataset show that compared with a separate training cluster, Alchemist significantly increases training throughput by up to 1.72x, reduces up to 47% memory usage during training, and supports up to 2x more training tokens - all while maintaining negligible impact on serving latency.
User modeling in large e-commerce platforms aims to optimize user experiences by incorporating various customer activities. Traditional models targeting a single task often focus on specific business metrics, neglecting the comprehensive user behavior, and thus limiting their effectiveness. To develop more generalized user representations, some existing work adopts Multi-task Learning (MTL)approaches. But they all face the challenges of optimization imbalance and inefficiency in adapting to new tasks. Continual Learning (CL), which allows models to learn new tasks incrementally and independently, has emerged as a solution to MTL's limitations. However, CL faces the challenge of catastrophic forgetting, where previously learned knowledge is lost when the model is learning the new task. Inspired by the success of prompt tuning in Pretrained Language Models (PLMs), we propose PCL, a Prompt-based Continual Learning framework for user modeling, which utilizes position-wise prompts as external memory for each task, preserving knowledge and mitigating catastrophic forgetting. Additionally, we design contextual prompts to capture and leverage inter-task relationships during prompt tuning. We conduct extensive experiments on real-world datasets to demonstrate PCL's effectiveness.
With the rise of generative models, there is a growing interest in unifying all tasks within a generative framework. Anomaly detection methods also fall into this scope and utilize diffusion models to generate or reconstruct normal samples when given arbitrary anomaly images. However, our study found that the diffusion model suffers from severe ``faithfulness hallucination'' and ``catastrophic forgetting'', which can't meet the unpredictable pattern increments. To mitigate the above problems, we propose a continual diffusion model that uses gradient projection to achieve stable continual learning. Gradient projection deploys a regularization on the model updating by modifying the gradient towards the direction protecting the learned knowledge. But as a double-edged sword, it also requires huge memory costs brought by the Markov process. Hence, we propose an iterative singular value decomposition method based on the transitive property of linear representation, which consumes tiny memory and incurs almost no performance loss. Finally, considering the risk of ``over-fitting'' to normal images of the diffusion model, we propose an anomaly-masked network to enhance the condition mechanism of the diffusion model. For continual anomaly detection, ours achieves first place in 17/18 settings on MVTec and VisA. Code is available at https://github.com/FuNz-0/One-for-More
Open-World Continual Learning (OWCL) is a challenging paradigm where models must incrementally learn new knowledge without forgetting while operating under an open-world assumption. This requires handling incomplete training data and recognizing unknown samples during inference. However, existing OWCL methods often treat open detection and continual learning as separate tasks, limiting their ability to integrate open-set detection and incremental classification in OWCL. Moreover, current approaches primarily focus on transferring knowledge from known samples, neglecting the insights derived from unknown/open samples. To address these limitations, we formalize four distinct OWCL scenarios and conduct comprehensive empirical experiments to explore potential challenges in OWCL. Our findings reveal a significant interplay between the open detection of unknowns and incremental classification of knowns, challenging a widely held assumption that unknown detection and known classification are orthogonal processes. Building on our insights, we propose \textbf{HoliTrans} (Holistic Knowns-Unknowns Knowledge Transfer), a novel OWCL framework that integrates nonlinear random projection (NRP) to create a more linearly separable embedding space and distribution-aware prototypes (DAPs) to construct an adaptive knowledge space. Particularly, our HoliTrans effectively supports knowledge transfer for both known and unknown samples while dynamically updating representations of open samples during OWCL. Extensive experiments across various OWCL scenarios demonstrate that HoliTrans outperforms 22 competitive baselines, bridging the gap between OWCL theory and practice and providing a robust, scalable framework for advancing open-world learning paradigms.
Large Language Models (LLMs) exhibit potential artificial generic intelligence recently, however, their usage is costly with high response latency. Given mixed LLMs with their own strengths and weaknesses, LLM routing aims to identify the most suitable model for each query in the stream to maximize response quality and minimize cost and latency. However, the challenges involve: (1) dynamic trade-offs among quality, cost, and latency; (2) enabling continual learning in deployed systems; and (3) navigating a varying (e.g., new LLM addition or old LLM removal) set of LLM candidates over time. To bridge these gaps, we develop MixLLM, a dynamic contextual-bandit-based routing system for query-LLM assignment. Specifically, we first leverage query tags to enhance query embeddings for the routing task. Next, we design lightweight prediction models to estimate the response qualities and costs of queries over LLMs. We then devise a meta-decision maker to choose the query-LLM assignments to best tradeoff response quality, cost, and latency. Finally, the system benefits from continual training, allowing it to adapt to evolving queries and user feedback over time. Our extensive experiments show that MixLLM achieves the best trade-offs in response quality, cost, and latency (97.25% of GPT-4's quality at 24.18% of the cost under the time constraint).
Low-Rank Adaptation (LoRA) is an efficient fine-tuning method that has been extensively applied in areas such as natural language processing and computer vision. Existing LoRA fine-tuning approaches excel in static environments but struggle in dynamic learning due to reliance on multiple adapter modules, increasing overhead and complicating inference. We propose Continual Low-Rank Adaptation (C-LoRA), a novel extension of LoRA for continual learning. C-LoRA uses a learnable routing matrix to dynamically manage parameter updates across tasks, ensuring efficient reuse of learned subspaces while enforcing orthogonality to minimize interference and forgetting. Unlike existing approaches that require separate adapters for each task, C-LoRA enables a integrated approach for task adaptation, achieving both scalability and parameter efficiency in sequential learning scenarios. C-LoRA achieves state-of-the-art accuracy and parameter efficiency on benchmarks while providing theoretical insights into its routing matrix's role in retaining and transferring knowledge, establishing a scalable framework for continual learning.
Online continual learning (OCL), which enables AI systems to adaptively learn from non-stationary data streams, is commonly achieved using experience replay (ER)-based methods that retain knowledge by replaying stored past during training. However, these methods face challenges of prediction bias, stemming from deviations in parameter update directions during task transitions. This paper identifies parameter variation imbalance as a critical factor contributing to prediction bias in ER-based OCL. Specifically, using the proposed parameter variation evaluation method, we highlight two types of imbalance: correlation-induced imbalance, where certain parameters are disproportionately updated across tasks, and layer-wise imbalance, where output layer parameters update faster than those in preceding layers. To mitigate the above imbalances, we propose the Parameter Variation Balancing Framework (PVBF), which incorporates: 1) a novel method to compute parameter correlations with previous tasks based on parameter variations, 2) an encourage-and-consolidate (E&C) method utilizing parameter correlations to perform gradient adjustments across all parameters during training, 3) a dual-layer copy weights with reinit (D-CWR) strategy to slowly update output layer parameters for frequently occuring sample categories. Experiments on short and long task sequences demonstrate that PVBF significantly reduces prediction bias and improves OCL performance, achieving up to 47\% higher accuracy compared to existing ER-based methods.
This literature review explores continual learning methods for on-device training in the context of neural networks (NNs) and decision trees (DTs) for classification tasks on smart environments. We highlight key constraints, such as data architecture (batch vs. stream) and network capacity (cloud vs. edge), which impact TinyML algorithm design, due to the uncontrolled natural arrival of data streams. The survey details the challenges of deploying deep learners on resource-constrained edge devices, including catastrophic forgetting, data inefficiency, and the difficulty of handling IoT tabular data in open-world settings. While decision trees are more memory-efficient for on-device training, they are limited in expressiveness, requiring dynamic adaptations, like pruning and meta-learning, to handle complex patterns and concept drifts. We emphasize the importance of multi-criteria performance evaluation tailored to edge applications, which assess both output-based and internal representation metrics. The key challenge lies in integrating these building blocks into autonomous online systems, taking into account stability-plasticity trade-offs, forward-backward transfer, and model convergence.
With the rapid development of large language models (LLMs), fully fine-tuning (FT) these models has become increasingly impractical due to the high computational demands. Additionally, FT can lead to catastrophic forgetting. As an alternative, Low-Rank Adaptation (LoRA) has been proposed, which fine-tunes only a small subset of parameters, achieving similar performance to FT while significantly reducing resource requirements. However, since LoRA inherits FT's design, the issue of catastrophic forgetting remains. To address these challenges, we propose SECURA: Sigmoid-Enhanced CUR Decomposition LoRA, a novel parameter-efficient fine-tuning (PEFT) variant that mitigates catastrophic forgetting while improving fine-tuning performance. Our method introduces a new normalization technique, SigNorm, to enhance parameter retention and overall performance. SECURA has been evaluated on a variety of tasks, including mathematical problem-solving (GSM8K), challenging question-answering (CNNDM), translation (NewsDE), and complex multiple-choice reasoning (LogiQA). Experimental results show that SECURA achieves an average fine-tuning improvement of 3.59% across four multiple-choice question (MCQ) tasks and a 2.51% improvement across five question-answering (QA) tasks on models such as Gemma2 2b, Qwen2 1.5b, Qwen 2 7b, Llama3 8b, and Llama3.1 8b, compared to DoRA. Moreover, SECURA demonstrates superior knowledge retention capabilities, maintaining more than 70% accuracy on basic LLM knowledge across 16 continual learning tests, outperforming Experience Replay (ER), Sequential Learning (SEQ), EWC, I-LoRA, and CUR-LoRA.
Continual learning (CL) is crucial for deploying large language models (LLMs) in dynamic real-world environments without costly retraining. While recent model ensemble and model merging methods guided by parameter importance have gained popularity, they often struggle to balance knowledge transfer and forgetting, mainly due to the reliance on static importance estimates during sequential training. In this paper, we present Recurrent-KIF, a novel CL framework for Recurrent Knowledge Identification and Fusion, which enables dynamic estimation of parameter importance distributions to enhance knowledge transfer. Inspired by human continual learning, Recurrent-KIF employs an inner loop that rapidly adapts to new tasks while identifying important parameters, coupled with an outer loop that globally manages the fusion of new and historical knowledge through redundant knowledge pruning and key knowledge merging. These inner-outer loops iteratively perform multiple rounds of fusion, allowing Recurrent-KIF to leverage intermediate training information and adaptively adjust fusion strategies based on evolving importance distributions. Extensive experiments on two CL benchmarks with various model sizes (from 770M to 13B) demonstrate that Recurrent-KIF effectively mitigates catastrophic forgetting and enhances knowledge transfer.
Online continuous action recognition has emerged as a critical research area due to its practical implications in real-world applications, such as human-computer interaction, healthcare, and robotics. Among various modalities, skeleton-based approaches have gained significant popularity, demonstrating their effectiveness in capturing 3D temporal data while ensuring robustness to environmental variations. However, most existing works focus on segment-based recognition, making them unsuitable for real-time, continuous recognition scenarios. In this paper, we propose a novel online recognition system designed for real-time skeleton sequence streaming. Our approach leverages a hybrid architecture combining Spatial Graph Convolutional Networks (S-GCN) for spatial feature extraction and a Transformer-based Graph Encoder (TGE) for capturing temporal dependencies across frames. Additionally, we introduce a continual learning mechanism to enhance model adaptability to evolving data distributions, ensuring robust recognition in dynamic environments. We evaluate our method on the SHREC'21 benchmark dataset, demonstrating its superior performance in online hand gesture recognition. Our approach not only achieves state-of-the-art accuracy but also significantly reduces false positive rates, making it a compelling solution for real-time applications. The proposed system can be seamlessly integrated into various domains, including human-robot collaboration and assistive technologies, where natural and intuitive interaction is crucial.
While 3D instance segmentation (3DIS) has advanced significantly, existing methods typically assume that all object classes are known in advance and are uniformly distributed. However, this assumption is unrealistic in dynamic, real-world environments where new classes emerge gradually and exhibit natural imbalance. Although some approaches have addressed class emergence, they often overlook class imbalance, resulting in suboptimal performance -- particularly on rare categories. To tackle this challenge, we propose CLIMB-3D, a unified framework for \textbf{CL}ass-incremental \textbf{Imb}alance-aware \textbf{3D}IS. Building upon established exemplar replay (ER) strategies, we show that ER alone is insufficient to achieve robust performance under constrained memory conditions. To mitigate this, we introduce a novel pseudo-label generator (PLG) that extends supervision to previously learned categories by leveraging predictions from a frozen prior model. Despite its promise, PLG tends to bias towards frequent classes. Therefore, we propose a class-balanced re-weighting (CBR) scheme, that estimates object frequencies from pseudo-labels and dynamically adjusts training bias -- without requiring access to past data. We design and evaluate three incremental scenarios for 3DIS on the challenging ScanNet200 dataset, and additionally on semantic segmentation on ScanNetV2. Our approach achieves state-of-the-art results, surpassing prior work by up to 16.76\% mAP for instance segmentation and approximately 30\% mIoU for semantic segmentation, demonstrating strong generalization across both frequent and rare classes.
Reinforcement learning (RL) algorithms have been successfully applied to control tasks associated with unmanned aerial vehicles and robotics. In recent years, safe RL has been proposed to allow the safe execution of RL algorithms in industrial and mission-critical systems that operate in closed loops. However, if the system operating conditions change, such as when an unknown fault occurs in the system, typical safe RL algorithms are unable to adapt while retaining past knowledge. Continual reinforcement learning algorithms have been proposed to address this issue. However, the impact of continual adaptation on the system's safety is an understudied problem. In this paper, we study the intersection of safe and continual RL. First, we empirically demonstrate that a popular continual RL algorithm, online elastic weight consolidation, is unable to satisfy safety constraints in non-linear systems subject to varying operating conditions. Specifically, we study the MuJoCo HalfCheetah and Ant environments with velocity constraints and sudden joint loss non-stationarity. Then, we show that an agent trained using constrained policy optimization, a safe RL algorithm, experiences catastrophic forgetting in continual learning settings. With this in mind, we explore a simple reward-shaping method to ensure that elastic weight consolidation prioritizes remembering both safety and task performance for safety-constrained, non-linear, and non-stationary dynamical systems.
Recent methods have made significant progress in synthesizing novel views with long video sequences. This paper proposes a highly scalable method for dynamic novel view synthesis with continual learning. We leverage the 3D Gaussians to represent the scene and a low-rank adaptation-based deformation model to capture the dynamic scene changes. Our method continuously reconstructs the dynamics with chunks of video frames, reduces the streaming bandwidth by $90\%$ while maintaining high rendering quality comparable to the off-line SOTA methods.
Typically, Few-shot Continual Relation Extraction (FCRE) models must balance retaining prior knowledge while adapting to new tasks with extremely limited data. However, real-world scenarios may also involve unseen or undetermined relations that existing methods still struggle to handle. To address these challenges, we propose a novel approach that leverages the Open Information Extraction concept of Knowledge Graph Construction (KGC). Our method not only exposes models to all possible pairs of relations, including determined and undetermined labels not available in the training set, but also enriches model knowledge with diverse relation descriptions, thereby enhancing knowledge retention and adaptability in this challenging scenario. In the perspective of KGC, this is the first work explored in the setting of Continual Learning, allowing efficient expansion of the graph as the data evolves. Experimental results demonstrate our superior performance compared to other state-of-the-art FCRE baselines, as well as the efficiency in handling dynamic graph construction in this setting.
6G networks aim to achieve global coverage, massive connectivity, and ultra-stringent requirements. Space-Air-Ground Integrated Networks (SAGINs) and Semantic Communication (SemCom) are essential for realizing these goals, yet they introduce considerable complexity in resource orchestration. Drawing inspiration from research in robotics, a viable solution to manage this complexity is the application of Large Language Models (LLMs). Although the use of LLMs in network orchestration has recently gained attention, existing solutions have not sufficiently addressed LLM hallucinations or their adaptation to network dynamics. To address this gap, this paper proposes a framework called Autonomous Reinforcement Coordination (ARC) for a SemCom-enabled SAGIN. This framework employs an LLM-based Retrieval-Augmented Generator (RAG) monitors services, users, and resources and processes the collected data, while a Hierarchical Action Planner (HAP) orchestrates resources. ARC decomposes orchestration into two tiers, utilizing LLMs for high-level planning and Reinforcement Learning (RL) agents for low-level decision-making, in alignment with the Mixture of Experts (MoE) concept. The LLMs utilize Chain-of-Thought (CoT) reasoning for few-shot learning, empowered by contrastive learning, while the RL agents employ replay buffer management for continual learning, thereby achieving efficiency, accuracy, and adaptability. Simulations are provided to demonstrate the effectiveness of ARC, along with a comprehensive discussion on potential future research directions to enhance and upgrade ARC.
Deep learning methods for material property prediction have been widely explored to advance materials discovery. However, the prevailing pre-train then fine-tune paradigm often fails to address the inherent diversity and disparity of material tasks. To overcome these challenges, we introduce MoMa, a Modular framework for Materials that first trains specialized modules across a wide range of tasks and then adaptively composes synergistic modules tailored to each downstream scenario. Evaluation across 17 datasets demonstrates the superiority of MoMa, with a substantial 14% average improvement over the strongest baseline. Few-shot and continual learning experiments further highlight MoMa's potential for real-world applications. Pioneering a new paradigm of modular material learning, MoMa will be open-sourced to foster broader community collaboration.
It is critical for robots to explore Federated Learning (FL) settings where several robots, deployed in parallel, can learn independently while also sharing their learning with each other. This collaborative learning in real-world environments requires social robots to adapt dynamically to changing and unpredictable situations and varying task settings. Our work contributes to addressing these challenges by exploring a simulated living room environment where robots need to learn the social appropriateness of their actions. First, we propose Federated Root (FedRoot) averaging, a novel weight aggregation strategy which disentangles feature learning across clients from individual task-based learning. Second, to adapt to challenging environments, we extend FedRoot to Federated Latent Generative Replay (FedLGR), a novel Federated Continual Learning (FCL) strategy that uses FedRoot-based weight aggregation and embeds each client with a generator model for pseudo-rehearsal of learnt feature embeddings to mitigate forgetting in a resource-efficient manner. Our results show that FedRoot-based methods offer competitive performance while also resulting in a sizeable reduction in resource consumption (up to 86% for CPU usage and up to 72% for GPU usage). Additionally, our results demonstrate that FedRoot-based FCL methods outperform other methods while also offering an efficient solution (up to 84% CPU and 92% GPU usage reduction), with FedLGR providing the best results across evaluations.
Our ability to continuously acquire, organize, and leverage knowledge is a key feature of human intelligence that AI systems must approximate to unlock their full potential. Given the challenges in continual learning with large language models (LLMs), retrieval-augmented generation (RAG) has become the dominant way to introduce new information. However, its reliance on vector retrieval hinders its ability to mimic the dynamic and interconnected nature of human long-term memory. Recent RAG approaches augment vector embeddings with various structures like knowledge graphs to address some of these gaps, namely sense-making and associativity. However, their performance on more basic factual memory tasks drops considerably below standard RAG. We address this unintended deterioration and propose HippoRAG 2, a framework that outperforms standard RAG comprehensively on factual, sense-making, and associative memory tasks. HippoRAG 2 builds upon the Personalized PageRank algorithm used in HippoRAG and enhances it with deeper passage integration and more effective online use of an LLM. This combination pushes this RAG system closer to the effectiveness of human long-term memory, achieving a 7% improvement in associative memory tasks over the state-of-the-art embedding model while also exhibiting superior factual knowledge and sense-making memory capabilities. This work paves the way for non-parametric continual learning for LLMs. Our code and data will be released at https://github.com/OSU-NLP-Group/HippoRAG.