跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
开发与测试

libero-inference-time-scaling

Run the LIBERO inference-time scaling experiment: debug Long-Pro tasks at frozen library snapshots, evaluate token-budget checkpoints, and plot Pareto curves.

NVlabs · 已有 AI 审核记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add NVlabs/ASPIRE --skill libero-inference-time-scaling

技能说明


name: libero-inference-time-scaling description: "Run the LIBERO inference-time scaling experiment: debug Long-Pro tasks at frozen library snapshots, evaluate token-budget checkpoints, and plot Pareto curves."

LIBERO Inference-Time Scaling

Use this skill for Experiment 5. Start with INSTRUCTIONS.md.

Run Order

  1. Start with INSTRUCTIONS.md.
  2. Set up snapshot worktrees for the target library sizes.
  3. Follow main-agent-prompt.md for Stage 1 debug + Stage 2 held-out eval.
  4. Use subagent-prompt.md for each debug subagent.
  5. Follow token-scaling-coordinator.md to select token-budget checkpoints and run Stage 2 variants.
  6. Use clean-task-slate.md before rerunning a snapshot or task.

Files

FilePurpose
INSTRUCTIONS.mdEnd-to-end human/agent entrypoint
main-agent-prompt.mdCoordinator recipe for debug+held-out eval
subagent-prompt.mdPer-task Stage 1 debug subagent prompt
clean-task-slate.mdReset checklist before reruns
token-scaling-coordinator.mdToken-budget checkpoint eval recipe
../library-size-scaling/main-agent-prompt.mdRelated zero-shot snapshot eval flow