Event ·Multi-source

NVIDIA 发布 Skill2Env:将公开 Agent Skills 编译为 RL 环境生成近 8000 条训练任务

NVIDIA 的 Skill2Env 流水线把公开的 Agent Skills 转成可执行终端任务:Codex planner 读取 SKILL.md 并拆分工作流,Codex creator 为每个任务构建程序化测试和行为 rubric。从约 3400 个爬取的 Skills 中产出 7971 条任务,覆盖 13 个领域,软件工程占比不到四分之一,用 GPT-5.6 Sol 生成的 API 花费超过 9 万美元。Qwen3.8-27B 经过 300 步结果导向 RL 训练后,Terminal-Bench 2.1 成绩从 49.4% 提升到 54.1%,S2EBench pass@1 从 33.4% 提升到 37.7%。加入 rubric 奖励在基准上增益较小(Terminal-Bench 2.1 为 50.1%),但人类偏好的模型轨迹占比从 54.5% 升至 73.0%。

Current read

NVIDIA 把公开 Agent Skills 自动转成 RL 训练环境,产出近 8000 条任务,还开源了流水线思路,做 Agent 训练的可以看看。

9 sources73° AI heatLast updated 2026/9/23 20:28:01

Evidence

9 sources
Related source 1shao__meng
View source
Related source 3arXiv cs.LG
View source
Related source 5rohanpaul_ai
View source
Related source 6Clement Delangue
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.