界面已本地化;技能介绍与说明文档按所选语言由 AI 自动翻译,代码、命令、路径和专业术语保留原文,以原作者文档为准。
探索技能
效率与协作 · SKILL.md

autovp_tuning_benchmark

autovp_tuning_benchmark

【效率与协作】Select AutoVP candidate configurations and record benchmark-style accuracy targets for recovery runs.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准

技能说明

GitHub

name: autovp_tuning_benchmark description: Select AutoVP candidate configurations and record benchmark-style accuracy targets for recovery runs.

AutoVP Tuning Benchmark

Use this skill when comparing AutoVP candidate configurations or writing benchmark result metadata. It should not claim full Table 2 reproduction for a reduced proxy.

Inputs

Candidate configurations, early validation metrics, dataset name, metric name, and paper target.

Outputs

Selected configuration and benchmark summary with proxy/full status.

Workflow

Normalize records, sort by metric with deterministic tie-breaking, choose the best candidate, and emit dataset/metric/target metadata. Preserve module-plan target values exactly.

Validation

Run included candidate-selection test.

我的收藏