Skill-RSI：把机器人操作技能“编译”成可复用代码的离线方法
Compile Once, Run Cheap: Frozen, Portable Skills for Embodied Agents, and What Recursive Self-Improvement Adds
这篇论文把机器人技能当代码来“编译”：Claude Code 离线写技能，冻结后跑起来一次不到半美分，还给了严格的实验协议，做具身智能的值得看看细节。
arXiv 论文提出 Skill-RSI，让 Claude Code 作为离线作者在仿真中编写、诊断并修复机器人操作技能，直到通过隐藏布局的考试后冻结成纯代码。在 RoboTwin 2.0 的 7 个杂乱桌面任务上，4 个任务成功编译为 7 个冻结技能，每个作者成本 4.29-8.22 美元；另外 3 个任务在 18 次尝试（含 5 次 GPT-6）中未能通过 95% 成功率、零安全中止的门槛。继承式修复在 8 个对照实验中全部胜过从零重写（sign test p = 0.004）。冻结技能在新布局上无模型情况下取得 148/150 成功，用 DeepSeek-flash 调用时每次运行仅 0.00028-0.00054 美元。
Compile Once, Run Cheap: Frozen, Portable Skills for Embodied Agents, and What Recursive Self-Improvement Adds
Embodied AI lacks the data-rich pretraining that gave language models their agents and skills, so new manipulation capability costs demonstrations per task or a frontier model in every episode. We study a third route: a strong general model acts only as an offline author that writes, runs, diagnoses and repairs a manipulation skill in simulation until it passes an exam on layouts drawn after freezing; the frozen skill then runs as code alone or under a cheap caller that reads only its description. We call this skill-level recursive self-improvement Skill-RSI and specify a compile protocol (isolated, audited author; hidden validation layouts; no-feedback and inherit-only controls; tamper-evident ledger) and a reuse protocol that runs the identical files with executors from another model family. On 7 RoboTwin 2.0 tasks with cluttered tables, Claude Code compiled 4 tasks into 7 frozen skills, 5 from a first draft needing only qualification, at USD 4.29-8.22 of author cost each (USD 55 per skill counting the 13 failed campaigns). The other three tasks did not compile in 18 campaigns, including five with GPT-6 and a pre-registered 150-layout run: on the hardest task every campaign scored 78-95 of 100 but had 3-16 safety aborts, failing the gate of 95% success with zero aborts; a 100-episode exam bounds an abort rate only to 3.6%. On hidden layouts, repair that inherits the previous version beat from-scratch rewriting in 8 of 8 controlled campaigns (sign test p = 0.004); an inherit-only lineage attributes most of the gap to inheritance, with development feedback adding a little (3 of 3, p = 0.125). Write-once campaigns found no detectable first-draft effect of practice (minimum detectable effect 14 points). The frozen skills kept 148/150 successes on new layouts with no model and 147/150 with a DeepSeek-flash caller at USD 0.00028-0.00054 per episode; no direct-control baseline was run.