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 2510.01239cs.CL
收藏

CIFLEX:面向单设备端LLM多轮交互中子任务执行的上下文指令流

CIFLEX: Contextual Instruction Flow for Sub-task Execution in Multi-Turn Interactions with a Single On-Device LLM

Qualcomm AI Research(高通人工智能研究院)

Juntae Lee, Jihwan Bang, Seunghan Yang, Simyung Chang

AI总结:

针对单设备端LLM多轮交互中子任务切换计算开销高的问题,提出CIFLEX系统,通过复用主任务KV缓存、侧路径注入指令及分层分类子任务选择策略,在不降低性能的前提下显著减少计算量,实现高效的设备端多任务对话。

摘要:

我们提出了CIFLEX(上下文指令流子任务执行系统),这是一种新型执行系统,用于在单设备端大语言模型(LLM)的多轮交互中高效处理子任务。随着LLM能力不断增强,人们期望单个模型能处理各类子任务,更有效、更全面地支持响应用户请求。朴素方法在主任务与子任务(如查询重写、摘要生成)间切换时会重新处理整个对话上下文,带来显著计算开销。CIFLEX通过复用主任务的键值(KV)缓存,仅将任务特定指令注入隔离的侧路径,从而降低这一开销。子任务执行完成后,模型通过缓存的上下文回退到主路径,避免了冗余的预填充计算。为支持子任务选择,我们还开发了专为小规模模型定制的分层分类策略,将多选决策分解为二元决策。实验表明,CIFLEX在不降低任务性能的前提下显著减少了计算成本,实现了设备端可扩展、高效的多任务对话。

英文摘要:

We present CIFLEX (Contextual Instruction Flow for Sub-task Execution), which is a novel execution system for efficient sub-task handling in multi-turn interactions with a single on-device large language model (LLM). As LLMs become increasingly capable, a single model is expected to handle diverse sub-tasks that more effectively and comprehensively support answering user requests. Naive approach reprocesses the entire conversation context when switching between main and sub-tasks (e.g., query rewriting, summarization), incurring significant computational overhead. CIFLEX mitigates this overhead by reusing the key-value (KV) cache from the main task and injecting only task-specific instructions into isolated side paths. After sub-task execution, the model rolls back to the main path via cached context, thereby avoiding redundant prefill computation. To support sub-task selection, we also develop a hierarchical classification strategy tailored for small-scale models, decomposing multi-choice decisions into binary ones. Experiments show that CIFLEX significantly reduces computational costs without degrading task performance, enabling scalable and efficient multi-task dialogue on-device.

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。

↑