跳到正文
返回论文列表
cs.CL提交于 已译

AgentCollab:面向高效 LLM 智能体的自评估驱动协作范式

AgentCollab: A Self-Evaluation-Driven Collaboration Paradigm for Efficient LLM Agents

Wenbo Gao · Renxi Liu · Xian Wang · Fang Guo · Shuai Yang · Xi Chen · et al.

中文摘要

由大语言模型(LLM)驱动的自主智能体通过长程推理与工具交互完成复杂任务,其中执行效率与推理鲁棒性之间存在根本性权衡。不同能力-成本层级的模型具有互补优势:低成本模型可快速执行,但在困难推理片段上可能表现不佳;更强模型以更高计算成本提供更稳健的推理。AgentCollab 是一个自驱动的协作推理框架,在智能体执行过程中动态协调不同推理能力的模型。该框架不依赖外部路由模块,而是利用智能体自身的自反思信号判断当前推理轨迹是否取得有意义进展,仅在必要时将控制权升级到更强的推理层级。为进一步稳定长程执行,引入难度感知的累积升级策略,根据近期失败信号分配额外推理预算。实验采用两级小-大模型设置实例化该框架。在多个多步智能体基准上的实验表明,AgentCollab 持续改善 LLM 智能体的准确率-效率帕累托前沿。代码见 https://github.com/SparkEngineAI/AgentFlashInfer。

关键要点

  1. 01问题:LLM 智能体长程执行中效率与推理鲁棒性存在根本权衡,不同能力-成本模型各有短板。
  2. 02方法:用智能体自身自反思信号判断推理进展,仅在必要时升级到更强推理层级,无需外部路由模块。
  3. 03方法:难度感知累积升级策略依据近期失败信号分配额外推理预算,稳定长程执行。
  4. 04结果:在多个多步智能体基准上,两级小-大模型设置持续改善准确率-效率帕累托前沿。

解读

尚无解读。

原始英文摘要

arXiv:2603.26034v2 Announce Type: replace Abstract: Autonomous agents powered by large language models (LLMs) perform complex tasks through long-horizon reasoning and tool interaction, where a fundamental trade-off arises between execution efficiency and reasoning robustness. Models at different capability-cost levels offer complementary advantages: lower-cost models enable fast execution but may struggle on difficult reasoning segments, while stronger models provide more robust reasoning at higher computational cost. We present AgentCollab, a self-driven collaborative inference framework that dynamically coordinates models with different reasoning capacities during agent execution. Instead of relying on external routing modules, the framework uses the agent's own self-reflection signal to determine whether the current reasoning trajectory is making meaningful progress, and escalates control to a stronger reasoning tier only when necessary. To further stabilize long-horizon execution, we introduce a difficulty-aware cumulative escalation strategy that allocates additional reasoning budget based on recent failure signals. In our experiments, we instantiate this framework using a two-level small-large model setting. Experiments on diverse multi-step agent benchmarks show that AgentCollab consistently improves the accuracy-efficiency Pareto frontier of LLM agents. The code is available at https://github.com/SparkEngineAI/AgentFlashInfer.

同方向论文 · cs.CL

查看全部 →

反向发现 · 这篇论文相关的工具与提示词

去交叉发现 →

关联工具 · 1

Google Research88

Google 将 Antigravity 的 /teamwork 多智能体工作流开放给外部研究者

Google 在 Antigravity 中开放 /teamwork 多智能体工作流及 /boost、Deep Review 等工具,外部研究者现在可以用 Gemini 4 模型调用。

推荐理由:Google 公开了 /teamwork 多智能体工作流在数学证明、化学结构解析和内核优化上的具体结果,并开放给外部研究者使用。