Search papers, labs, and topics across Lattice.
This paper introduces the FiUni framework, which integrates batch-level task detection with parameter-efficient continual adaptation to address catastrophic forgetting in large language models (LLMs). By leveraging the Fisher information matrix (FIM) to identify task similarities without explicit boundaries, FiUni dynamically adjusts its learning strategy to either reuse knowledge or create new subspaces as needed. Experimental results demonstrate that FiUni achieves competitive performance against state-of-the-art task-aware continual learning methods while utilizing fewer trainable parameters.
FiUni reveals that task-free continual learning can be achieved by intelligently leveraging Fisher information, enabling LLMs to adapt without explicit task boundaries.
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.
