Large language models (LLMs) can adapt to new tasks via in-context learning (ICL) without parameter updates, making them powerful learning engines for fast adaptation. While extensive research has examined ICL as a few-shot learner, whether it can achieve long-term retention and cross-task knowledge accumulation when multitasks arrive sequentially remains underexplored. Motivated by human memory studies, we investigate the retention characteristics of ICL in multitask settings and extend it to in-context continual learning (ICCL), where continual learning ability emerges through task scheduling and prompt rearrangement. Experiments on Markov-Chain benchmarks demonstrate that, for specific large-language models, ICCL benefits from distributed practice (DP) in a manner analogous to humans, consistently revealing a spacing "sweet spot" for retention. Beyond retention performance, we propose a human-retention similarity metric to quantify how closely a continual-learning (CL) method aligns with human retention dynamics. Using this metric, we show that linear-attention models such as MAMBA and RWKV exhibit particularly human-like retention patterns, despite their retention performance lagging behind that of Transformer-based LLMs. Overall, our results establish ICCL as both cognitively plausible and practically effective, providing an inference-only CL paradigm that mitigates catastrophic forgetting and addresses the stability-plasticity dilemma in conventional CL methods.
Building similarity graph...
Analyzing shared references across papers
Loading...
Liuwang Kang
Institute of Art
Fan Wang
China Tourism Academy
Shaoshan Liu
Shenzhen Academy of Robotics
Building similarity graph...
Analyzing shared references across papers
Loading...
Kang et al. (Fri,) studied this question.
synapsesocial.com/papers/68f6196ee0bbbc94fac3639e — DOI: https://doi.org/10.48550/arxiv.2509.22764
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: