Experiments show that FiUni can effectively infer latent batch-level task affiliations and achieve competitive performance against advanced task-aware CL methods with fewer trainable parameters.
Abstract
To mitigate catastrophic forgetting in downstream continual learning (CL) for large language models (LLMs), existing methods typically constrain parameter updates or introduce task-specific adaptation modules. However, these methods often rely on explicit task boundaries during training, limiting their applicability to realistic task-free scenarios. In this paper, we propose a \textbf{Fi}sher-guided \textbf{uni}fied (\textbf{FiUni}) framework for batch-level task detection and parameter-efficient continual adaptation. FiUni is motivated by a key observation about the Fisher information matrix (FIM) of pre-trained models: the orthogonality among the principal subspaces of its Kronecker-Factored Approximate Curvature (K-FAC) approximation, estimated from a small number of downstream task samples, can reflect the similarity between different tasks. Based on this observation, FiUni constructs FIM-derived frozen subspaces to guide low-rank adaptation (LoRA), while matching the Fisher principal subspace of each incoming batch window with historical subspaces. This enables FiUni to adaptively determine whether to reuse existing knowledge, expand a related subspace, or create a new subspace, dynamically balancing knowledge sharing and task isolation. Experiments show that FiUni can effectively infer latent batch-level task affiliations and achieve competitive performance against advanced task-aware CL methods with fewer trainable parameters.
This work observes that pooled token embeddings from a frozen LLM embedding layer already separate task distributions throughout the learning sequence, and concludes that a Gaussian mixture model fitted on these embeddings, without any gradient-based training, is sufficient for task-agnostic adapter selection at test time, eliminating the need for a learned gating module.
TASSO, a new paradigm that efficiently preserves the latent space geometry while ensuring network plasticity, is introduced with two complementary techniques: subspace learning and geometry-aware knowledge distillation.
Changming Sun, Francesco Barbato, Matteo Caligiuri et al.· 0 citations
This work introduces a novel measure, the relative parameter-importance, which measures the relative importance of each parameter with respect to both the current and past tasks, and allows the update of parameters with high past-task importance when they have low relative-importance, thus enabling backward knowledge transfer.
Malavika Suresh, I. Nkisi-Orji, N. Wiratunga· 0 citations
Low-Rank Adaptation (LoRA) enables foundation models to efficiently adapt to downstream tasks with minimal trainable parameters, yet it lacks the capability of continual learning (CL) from non-stationary data streams. While existing CL methods attempt to address this by storing and retrieving task-specific parameter modules, they suffer from non-trivial challenges of task identification and inconsistency in input-output feature mappings, resulting in suboptimal feature representations. Instead of keeping multiple task-specific parameter modules, we propose a novel continual LoRA mechanism that parameterizes all task-specific adaptations with a single adaptation module, naturally obviating the task identification and mapping inconsistency issues. To this end, we reformulate LoRA-based CL as a consistent feature mapping problem that mimics the behavior of the joint-training upper bound, wherein a unified adaptation parameter matrix is learned to simultaneously capture the input-output relationships established by all task-specific LoRAs. By solving this formulation, our approach facilitates approximate unified optimization over sequential tasks via progressively updating cumulative feature statistics, achieving efficient knowledge retention and renewal. Notably, the proposed arithmetic training mechanism can be applied both to the backbone of the foundation model for parameter-efficient fine-tuning and the classifier optimization, thereby enabling continual learning within an integrated architecture that closely approximates the joint learning paradigm. Extensive experiments across conventional and long-sequence class-incremental benchmarks demonstrate that our method achieves new state-of-the-art performance.
Yue Lu, Shizhou Zhang, De Cheng et al.· IEEE Transactions on Pattern...· 0 citations
Vision-language models (VLMs), such as contrastive language-image pre-training (CLIP), exhibit powerful zero-shot generalization capabilities. Parameter-efficient fine-tuning (PEFT) techniques, notably prompt learning, have been extensively explored to adapt these models to downstream tasks. However, their efficacy remains constrained when transferred to specialized domains like remote sensing. We argue that the bottleneck stems not merely from the limited parameters of prompts, but essentially from the disruption of the input’s original image–text features and the lack of deep cross-modal alignment. In particular, existing methods typically rely on global attention or coarse-grained feature mapping. This inadvertently corrupts the original input representations, thereby impairing the model’s inherent generalization. Furthermore, their isolated unimodal gradient updates fail to bridge the semantic gap inherent in complex remote sensing scenes. To address these challenges, we propose tokenwise prompt-free learning (Tiper), shifting the optimization paradigm from introducing external prompts to precisely recalibrating the critical tokens that govern classification outputs. In particular, Tiper employs a hierarchical learner to supersede global prompts. Crucially, this learner intervenes exclusively on the specific core tokens (i.e., the CLS token in the visual branch and the EOT token in the textual branch), leaving other original input representations unperturbed. This fine-grained strategy effectively balances domain adaptation with the preservation of inherent generalization. Finally, we design the learner as a cross-modal coupled bridge with shared weights, enabling it to synchronously receive gradient feedback from both modalities and fostering profound multimodal collaboration. Extensive experiments validate our method on eight public remote sensing datasets covering diverse scenes and resolutions. In the base-to-new generalization task, Tiper outperforms the strong baseline MaPLe with a significant 3.7% improvement in the harmonic mean (HM). Notably, without relying on any external large-scale domain models, Tiper surpasses the latest domain-specific prompt learning methods (e.g., domain-controlled prompt learning (DCPL), domain prompt learning with quaternion networks (DPLQ)), demonstrating its superior adaptability for remote sensing image scene classification.
Tengfei Gong, Jun-Lin Wu, Yaxioong Chen et al.· IEEE Transactions on Geoscie...· 0 citations
TAILS resolves cross-task ambiguity at the representation level, while leaving the original PTM, method-specific modules, and classifier unchanged, and can improve classification and task-inference performance with modest parameter overhead and negligible inference cost.
Zhiming Xu, Huiyu Yi, Zhenghe Xie et al.· 0 citations
