bench
Run benchmark and analyze PDF parsing performance
📖 中文导读:Run benchmark and analyze PDF parsing performance 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: bench description: Run benchmark and analyze PDF parsing performance
/bench
Builds Java and runs the full benchmark suite, then analyzes results.
Execution Steps
- Run
./scripts/bench.sh - Extract metrics from
tests/benchmark/prediction/opendataloader/evaluation.jsonusing:jq '{summary, metrics, table_detection, speed}' tests/benchmark/prediction/opendataloader/evaluation.json - Output summary:
- NID (reading order)
- TEDS (table structure)
- MHS (heading structure)
- Table Detection F1/Precision/Recall
- Speed (s/doc, total time)
- Compare with thresholds in
tests/benchmark/thresholds.json - Warn if regression detected
Options
/bench --doc-id 01030000000189- Run for a specific document only
Notes
- Benchmark won't run if Java build fails
- First run installs Python dependencies via uv sync