arXiv：cs.AI· Xi Qin, Isabel Kurth, Xin Cui, Elin Park, Alexander Schaefer, Yaad Oren·· 5 小时前AI 评分42
用 Claude Opus 生成终端智能体训练任务为何失效：数据生成与验证挑战解析
When Terminal-Agent Training Stalls: Demystifying Data Generation and Verification Challenge
AI 导读
研究用 Claude Opus 作为元智能体生成终端任务与验证器用于 RL 训练，诊断出基准无效、执行框架脆弱、奖励错位三类失败。提示词重设计与上下文扩展将基线可解率提升 5.6 倍，但 9B 模型在 Claude Opus 生成任务上 20 步内 pass@2 均值饱和于 81.3%；加入困难任务后该指标降至 20.6%，训练配置未变，表明可解区间因模型而异。
正文
Abstract:Using a frontier model like Claude Opus as a meta-agent to generate terminal tasks and verifiers for RL training is increasingly common. Yet a runnable Docker image and executable test suite do not guarantee a faithful end-to-end pipeline for terminal agent training. We present a meta-agent pipeline motivated by this gap, diagnosing three classes of failure: benchmark invalidity, harness brittleness, and reward misalignment. Prompt redesign and context extension raise baseline solvability 5.6 times, but a 9B model saturates at 81.3% mean pass@2 within 20 steps on Claude Opus-generated tasks. Adding hard tasks reduces mean pass@2 to 20.6% without changing the training configuration, a strong evidence that the solvability band is model-specific. These findings demonstrate that meta-agent reliability requires solvability-band calibration, verifier audits, and infrastructure error accounting as first-class evaluation criteria, not post-hoc diagnost.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.02405 [cs.AI] |
| (or arXiv:2610.02405v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.02405 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Xi Qin [view email]
[v1]
Thu, 1 Oct 2026 19:32:44 UTC (81 KB)
来源：arXiv：cs.AI · arxiv.org