arXiv cs.CL· Saiyue Lyu, Jerrold Huang, Yunbei Zhang, Janet Wang, Zhenting Wang, Yingqiang Ge, Haibo Zhang·· 13 小时前AI 评分42
卡住的 Agent 下一步该把 token 花在哪？推理期资源的因果归因
Where Should a Stuck Agent Spend Its Next Tokens? Causal Attribution of Inference-Time Resources
AI 导读
研究者提出 Banzhaf Resource Attribution（BRA）框架，在 MBPP+、HumanEval+ 的 194 个函数级失败和 SWE-bench Verified 的 126 个仓库级失败上评估六个模型，量化计算、信息、反馈三类推理期资源的边际贡献。
正文 · 原文
Abstract:Inference-time scaling is often treated as a token-budget problem, but a failed coding agent can seek qualitatively different resources: computation, information, or feedback. We introduce Banzhaf Resource Attribution (BRA), a controlled factorial framework that measures each resource's marginal contribution to repair and its interactions while holding the failed state and revision protocol fixed. We evaluate six models on 194 function-level failures from MBPP+ and HumanEval+ and 126 repository-level failures from SWE-bench Verified. Task-relevant specifications consistently improve repair, whereas public-test feedback provides small gains and retrieved code little to none. The value of additional computation declines with model capability, while information remains useful; resource channels largely exhibit diminishing returns, although complementary information can reinforce itself. Yet three of four frontier models select feedback in 95-100% of cases and obtain lower gains than uniform random resource selection. BRA thus exposes a gap between the resources agents prefer and those that measurably improve repair.
| Subjects: | Software Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.28834 [cs.SE] |
| (or arXiv:2610.28834v1 [cs.SE] for this version) | |
| https://doi.org/10.48550/arXiv.2610.28834 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Saiyue Lyu [view email]
[v1]
Tue, 20 Oct 2026 18:59:01 UTC (211 KB)
来源：arXiv cs.CL · arxiv.org