arXiv：cs.CL· Jun Zhao, Leiming Fu, Yanbo Wen, Yiding Wang, Xuantong Liu, Yang Shu, Yuyang Lu, Xuanran Xing, Jingqi Tong, Hao Xu, Qi Zhang, Xuanjing Huang·· 3 小时前AI 评分41
LiveMACE：面向演化市场中 LLM 智能体能力的过程感知评测
LiveMACE: Process-Aware Evaluation of LLM Agent Capabilities in Evolving Markets
AI 导读
研究者提出 LiveMACEBench，一个以实时金融市场为动态测试床、针对持久 LLM 智能体的过程感知基准，5 个前沿 LLM 在匹配的 Tool Use、Persistent Memory、Rule Following 与 Multi-Agent Collaboration 配置下连续运行。
正文
Authors:Jun Zhao, Leiming Fu, Yanbo Wen, Yiding Wang, Xuantong Liu, Yang Shu, Yuyang Lu, Xuanran Xing, Jingqi Tong, Hao Xu, Qi Zhang, Xuanjing Huang
Abstract:Evaluating agents by outcomes alone can obscure the capabilities that produce them. This problem is especially pronounced in evolving environments, where outcomes reflect a closed-loop interaction between agent behavior and changing external conditions. We introduce LiveMACEBench, a process-aware benchmark that uses live financial markets as a naturally evolving testbed for persistent LLM agents. Five frontier LLMs operate along continuous trajectories under matched Tool Use, Persistent Memory, Rule Following, and Multi-Agent Collaboration configurations. We evaluate them through both realized outcomes and mechanism-specific diagnostics derived from complete decision traces. Across 30 days of live evaluation, we find a pronounced outcome-capability gap: realized returns often diverge from capability-specific measurements, and similar outcomes can arise from markedly different patterns of mechanism use. Trace-level diagnostics further expose distinct bottlenecks across capabilities, demonstrating that mechanism access, effective mechanism use, and downstream performance are not interchangeable measures of agent capability. LiveMACEBench makes this distinction measurable, turning live markets from a performance leaderboard into a diagnostic environment for agent capability
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.09872 [cs.AI] |
| (or arXiv:2610.09872v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.09872 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jun Zhao [view email]
[v1]
Wed, 7 Oct 2026 11:33:21 UTC (6,239 KB)
来源：arXiv：cs.CL · arxiv.org