**摘要**
大型语言模型( LLM )通过更新其参数(例如,通过RL )为下游任务进行训练。然而,更新参数迫使他们吸收特定任务的信息,这可能导致灾难性的遗忘和可塑性丧失。相比之下,使用固定LLM参数的上下文学习可以便宜且快速地适应特定任务的需求(例如,提示优化) ,
👤 作者: Rishabh Tiwari, Kusha Sareen, Lakshya A Agrawal, Joseph E. Gonzalez, Matei Zaharia, Kurt Keutzer, Inderjit S Dhillon, Rishabh Agarwal, Devvrit Khatri
---
🔗 **[Learning, Fast and Slow: Towards LLMs That Adapt Continually](https://arxiv.org/abs/2605.12484v1)**
> Learning, Fast and Slow: Towards LLMs That Adapt Continually
🏷️ 来源: ArXiv cs.AI
⏱️ 2026-05-14 08:00
news
Learning, Fast and Slow: Towards LLMs That Adapt Continually
加载回复中...