# LoT H3 benchmarks

- [Experiment history: runs, results, limits, and what was tested](experiment-history.md)
- [Current r4 A/B summary](campaign-r4-results.json)
- [OpenVid step-0 layout comparison](openvid-pilot-step0-layout-screen.csv)
- [Runtime microbenchmarks](runtime-benchmarks.csv)
- [Step-0/250/500 OpenVid pilot CSVs](#pilot-csvs)
- [r4 control and preserve validation CSVs](#r4-validation-csvs)
- [r4 untouched confirmation CSVs](control-confirmation-step-500.csv) and [preserve](preserve-confirmation-step-500.csv)

## Pilot CSVs

- [Step 0 baseline](openvid-pilot-step-00000000.csv) and [aggregate JSON](openvid-pilot-step0-baseline.json)
- [Step 250](openvid-pilot-step-00000250.csv)
- [Step 500](openvid-pilot-step-00000500.csv)

## r4 validation CSVs

- [Control step 250](r4-control-validation-step-250.csv) · [control step 500](r4-control-validation-step-500.csv)
- [Preserve step 250](r4-preserve-validation-step-250.csv) · [preserve step 500](r4-preserve-validation-step-500.csv)

The r4 run ended without a promoted winner. Metric definitions, sample sizes, and limits are documented in the experiment history.