arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2409.18014cs.AI
收藏

Role-RL:基于角色强化学习的在线长上下文处理,使不同LLM在各自最优角色中发挥作用

Role-RL: Online Long-Context Processing with Role Reinforcement Learning for Distinct LLMs in Their Optimal Roles

South China Normal University(华南师范大学);Toronto University(多伦多大学);Zhejiang University(浙江大学)

Lewei He, Tianyu Shi, Pengran Huang, Bingzhi Chen, Qianglong Chen, Jiahui Pan

AI总结:

针对长上下文LLM的实现复杂性与模型选择困境,提出在线长上下文处理范式(OLP)及角色强化学习框架(Role-RL),根据实际性能自动部署不同LLM,在OLP-MINI数据集上实现93.2%平均召回率并节省79.4%成本。

摘要:

具有长上下文处理能力的大型语言模型(LLMs)在实现复杂性、训练效率和数据稀疏性方面仍面临挑战。为解决这一问题,我们提出了一种名为在线长上下文处理(OLP)的新范式,用于处理无限长度的文档,这通常出现在自动新闻报道、直播电商和爆款短视频等多种流媒体的信息接收与组织中。此外,当试图从爆炸式增长的众多LLMs中选择最合适的模型,以期获得卓越性能、低廉价格和短暂响应延迟时,我们经常面临两难境地。鉴于此,我们还开发了角色强化学习(Role-RL),根据不同LLMs的实际性能,在OLP流水线中自动将它们部署到各自的角色中。我们在OLP-MINI数据集上进行了大量实验,结果发现采用Role-RL框架的OLP达到了OLP基准,平均召回率为93.2%,并节省了79.4%的LLM成本。代码和数据集已公开于:https://anonymous.4open.science/r/Role-RL。

英文摘要:

Large language models (LLMs) with long-context processing are still challenging because of their implementation complexity, training efficiency and data sparsity. To address this issue, a new paradigm named Online Long-context Processing (OLP) is proposed when we process a document of unlimited length, which typically occurs in the information reception and organization of diverse streaming media such as automated news reporting, live e-commerce, and viral short videos. Moreover, a dilemma was often encountered when we tried to select the most suitable LLM from a large number of LLMs amidst explosive growth aiming for outstanding performance, affordable prices, and short response delays. In view of this, we also develop Role Reinforcement Learning (Role-RL) to automatically deploy different LLMs in their respective roles within the OLP pipeline according to their actual performance. Extensive experiments are conducted on our OLP-MINI dataset and it is found that OLP with Role-RL framework achieves OLP benchmark with an average recall rate of 93.2% and the LLM cost saved by 79.4%. The code and dataset are publicly available at: https://anonymous.4open.science/r/Role-RL.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑