用轻量框架实时纠错，让多智能体系统更省 token 且更可靠
Stop Wasting Your Tokens: Towards Efficient Runtime Multi-Agent Systems
- 通过无 LLM 的自适应过滤器实时触发干预
- 在 GAIA 基准上减少 29.68% token 消耗，成功率不变
- 适用于数学推理、代码生成等多场景，适配主流模型
多智能体系统虽擅长复杂任务，但随着自主性提升，操作复杂性常导致严重效率问题，如过度消耗 token 及因错误信息引发失败。现有方法多聚焦事后故障归因，缺乏主动、实时的干预机制以提升鲁棒性与效率。为此，我们提出 SupervisorAgent——一种轻量、模块化的运行时自适应监督框架，无需修改基础智能体结构即可工作。该框架由无 LLM 的自适应滤波器触发，在关键节点主动纠正错误、引导低效行为并净化观测数据。在挑战性基准 GAIA 上，SupervisorAgent 将 Smolagent 框架的 token 消耗平均降低 29.68%，且未影响其成功率。在数学推理、代码生成、问答等五个额外基准及多种 SOTA 基础模型上的广泛实验验证了方法的普适性与鲁棒性。
原文摘要 · Abstract (English)
While Multi-Agent Systems (MAS) excel at complex tasks, their growing autonomy with operational complexity often leads to critical inefficiencies, such as excessive token consumption and failures arising from misinformation. Existing methods primarily focus on post-hoc failure attribution, lacking proactive, real-time interventions to enhance robustness and efficiency. To this end, we introduce SupervisorAgent, a lightweight and modular framework for runtime, adaptive supervision that operates without altering the base agent's architecture. Triggered by an LLM-free adaptive filter, SupervisorAgent intervenes at critical junctures to proactively correct errors, guide inefficient behaviors, and purify observations. On the challenging GAIA benchmark, SupervisorAgent reduces the token consumption of the Smolagent framework by an average of 29.68% without compromising its success rate. Extensive experiments across five additional benchmarks (math reasoning, code generation, and question answering) and various SoTA foundation models validate the broad applicability and robustness of our approach.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。