Claude Code 等编码智能体的成本低效行为分析与缓解

View PDF HTML (experimental)

Abstract:Although effective, coding agents often incur substantial monetary costs. Their recurring cost-inefficient behaviors remain underexplored. We conduct the first study of behavioral cost inefficiencies in coding agents, analyzing 1,200 trajectories from Claude Code and Mini-SWE-Agent across four configurations on SWE-bench Verified. We identify three cost-inefficient behaviors: subsumed retrieval, similar script generation, and test re-execution. We then evaluate three mitigation strategies: structure-aware retrieval, agent-synthesized skills, and developer-designed skills, over 10k trajectories on held-out SWE-bench Verified and Pro tasks. Our main findings are: (1) The three behaviors affect 79.00%--98.00% of coding tasks and account for up to 22.75% of task cost. (2) Structure-aware retrieval can introduce retrieval overhead and alter agent delegation, causing inconsistent improvements in retrieval efficiency and cost increases of up to 28.14%. (3) Agent-synthesized skills tend to produce low-level, trace-specific guidance, limiting their effectiveness and generality. (4) In contrast, developer-designed skills provide high-level, trace-agnostic guidance, reducing cost by up to 41.73%, roughly twice the maximum gain from agent-synthesized skills.

Comments:	Under Review for Submission
Subjects:	Artificial Intelligence (cs.AI); Software Engineering (cs.SE)
Cite as:	arXiv:2609.30725 [cs.AI]
	(or arXiv:2609.30725v3 [cs.AI] for this version)
	https://doi.org/10.48550/arXiv.2609.30725 arXiv-issued DOI via DataCite

Submission history

From: Yiran Hu [view email]
[v1] Fri, 25 Sep 2026 02:54:18 UTC (797 KB)
[v2] Mon, 28 Sep 2026 02:40:04 UTC (799 KB)
[v3] Mon, 12 Oct 2026 04:31:43 UTC (790 KB)