analiza aiqbench
This research proposes SVC, a parameter-efficient method for continual learning in large language models that resolves the stability-plasticity dilemma by selectively updating singular-vector channels. By adaptively choosing which channels to train based on adaptation benefits and forgetting costs, the approach effectively mitigates catastrophic forgetting while maintaining strong downstream performance across multiple LLM families. The work advances the theoretical understanding of where plasticity should be allocated within transformer architectures to enable robust domain-specific adaptation without compromising pretraining knowledge.
Stability-Plasticity Balance via Singular-Vector Selection in LLM Continual Learning
arXiv:2610.11076v1 Announce Type: cross Abstract: Domain-specific continual adaptation of LLMs risks catastrophic forgetting, creating a fundamental tension between acquiring new capabilities and preserving those learned during pretraining. PEFT mitigates this problem by restricting the number of trainable parameters, but existing methods lack a principled unit for deciding where plasticity should be allocated and stability should be preserved. We identify the singular-vector channel as a natural unit for managing this trade-off. Each channel represents an input-output transformation, which can be updated to acquire new knowledge or fixed to preserve pretrained capabilities. Based on this perspective, we introduce SVC, a parameter-efficient continual-learning method that selectively updates Singular-Vector Channels. Before fine-tuning, SVC uses domain-specific data to estimate each channel's adaptation benefit and a fixed public general-domain corpus only as a history activation proxy for estimating forgetting cost. It then adaptively selects trainable channels based on these scores via knee-based cost screening, Pareto-front filtering, and Otsu thresholding. Experimental results across four LLM families and eight downstream tasks show that SVC better preserves pretrained capabilities while achieving strong downstream performance relative to existing PEFT baselines. Further analysis of channel scoring and selection demonstrates that selective plasticity at the singular-vector-channel level enables effective continual LLM adaptation.
Abstract:Domain-specific continual adaptation of LLMs risks catastrophic forgetting, creating a fundamental tension between acquiring new capabilities and preserving those learned during pretraining. PEFT mitigates this problem by restricting the number of trainable parameters, but existing methods lack a principled unit for deciding where plasticity should be allocated and stability should be preserved. We identify the singular-vector channel as a natural unit for managing this trade-off. Each channel represents an input-output transformation, which can be updated to acquire new knowledge or fixed to preserve pretrained capabilities. Based on this perspective, we introduce SVC, a parameter-efficient continual-learning method that selectively updates Singular-Vector Channels. Before fine-tuning, SVC uses domain-specific data to estimate each channel's adaptation benefit and a fixed public general-domain corpus only as a history activation proxy for estimating forgetting cost. It then adaptively selects trainable channels based on these scores via knee-based cost screening, Pareto-front filtering, and Otsu thresholding. Experimental results across four LLM families and eight downstream tasks show that SVC better preserves pretrained capabilities while achieving strong downstream performance relative to existing PEFT baselines. Further analysis of channel scoring and selection demonstrates that selective plasticity at the singular-vector-channel level enables effective continual LLM adaptation.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2610.11076 [cs.LG]
(or arXiv:2610.11076v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.11076 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 292 cuvinte
Subiecte conexe
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
LessWrong · 5 oct. 2026
Continual Reinforcement Learning with Neuroevolution
arXiv · 2 oct. 2026
Gated Adaptation for Continual Learning in Human Activity Recognition
arXiv · 31 iul. 2026
EvoBrain: Continual Learning of EEG Foundation Models Across Heterogeneous BCI Tasks
arXiv · 3 iun. 2026