Stanford 团队提出基准维护债概念并发布 Terminal-Bench 2.1
Benchmarks as Software: Terminal-Bench and the Case for Continuous Benchmarks
AI 导读
arXiv 论文提出基准维护债概念，并给出结合社区报告、oracle 检查与人工把关的持续验证流程。对经发布前审查的 Terminal-Bench 2.0 审计后，约 30% 任务存在缺陷，修复版 Terminal-Bench 2.1 使准确率提升超 10 个百分点，榜单名次最多变动三位但排序仍高度相关（Spearman rho = 0.958）。
正文 · 原文
Abstract:Coding agents are among the most economically consequential deployments of LLMs to date, making reliable evaluation increasingly important. Yet even benchmarks that are carefully validated at release can become stale. As models improve, they can discover valid solutions that existing verifiers reject, while changes to evaluation infrastructure can invalidate assumptions embedded in tasks and tests. We call the resulting accumulation of defects benchmark maintenance debt: a growing gap between benchmark scores and the capabilities they are intended to measure. We introduce a continuous-validation workflow for agentic benchmarks that combines community reports, task artifacts, agent traces, automated oracle checks, and human-gated repairs. We apply this workflow to Terminal-Bench 2.0, a benchmark that underwent extensive human and LM-assisted review and red-teaming before release. Our audit identifies issues in roughly 30% of the benchmark, and we release the resulting repairs as Terminal-Bench 2.1. These repairs change evaluation results: accuracy increased by over 10 percentage points and leaderboard positions move by up to three ranks, even as the overall ranking remains highly correlated (Spearman rho = 0.958).
We evaluate how resource allocation shapes benchmark conclusions. We show that fixed wall-clock can materially change agent performance: For example, extending the time limit closes a 20pp harness gap on GPT 5.4 and shrinks a 30 pp gap on GPT 5.4 mini to 5 pp. Benchmark maintenance is therefore not only about repairing task artifacts: resource budgets are part of the measurement specification and must be validated alongside the tasks themselves.
More broadly, benchmark maintenance debt is not just accumulated implementation bugs: our results show that as models improve, both evaluation assumptions and resource settings may need to be revalidated.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.08417 [cs.AI] |
| (or arXiv:2611.08417v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.08417 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Estefany Kelly Buchanan [view email]
[v1]
Thu, 5 Nov 2026 18:18:39 UTC (601 KB)
来源：arXiv cs.AI · arxiv.org