arXiv cs.CL· Hamish Ivison, Junjie Oscar Yin, Rulin Shao, Teng Xiao, Nathan Lambert, Hannaneh Hajishirzi·· 12 小时前AI 评分58
TMax：面向终端智能体的简单强化学习训练配方
Tmax: A simple recipe for terminal agents
AI 导读
TMax 是一套开源的终端智能体强化学习训练配方，仅用 9B 参数就在 Terminal-Bench 2.0 上取得 33%，超过此前更大模型的表现。该配方先用一套新分类法生成数据，结合难度控制、人物设定和验证器多样化，低成本产出大量终端环境用于 RL 和 SFT 训练，再以仅依赖结果、小幅调整稳定性的方式训练开放权重模型。方法适用于 2B 到 27B 的多个模型家族，作者开源了数据、模型和代码。
正文 · 原文
Abstract:Terminal-using agents have quickly become a popular and important downstream application of language models (LMs). Despite their prevalence, relatively little academic work has examined RL-based training of these models, due to difficult benchmarks, a lack of data, and a lack of simple baseline recipes. We present TMax, an open RL recipe for terminal agents that brings small open models closer to the frontier. While simple, our recipe achieves 33% on Terminal-Bench 2.0 with only 9B parameters, outperforming much larger models from prior work. Our recipe first involves generating data using a novel taxonomy, combining difficulty control, personas, and verifier diversification, which allows us to cheaply generate large amounts of terminal environments for RL and SFT training. We then train open-weight models using RL with our data, using a simple, outcome-only recipe, with small adjustments to improve stability. Our recipe works for models from 2B to 27B and across model families, dominating the Pareto curve for open terminal-agent performance at smaller sizes. We release our data, models, and code as a strong baseline for future open academic work on terminal agents at this https URL.
| Comments: | preprint |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2606.23321 [cs.CL] |
| (or arXiv:2606.23321v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2606.23321 arXiv-issued DOI via DataCite |
Submission history
From: Hamish Ivison [view email]
[v1]
Mon, 22 Jun 2026 13:32:52 UTC (3,023 KB)
[v2]
Mon, 2 Nov 2026 02:20:09 UTC (3,079 KB)
来源：arXiv cs.CL · arxiv.org