exp01-baseline

Which claim does this support?

If you cannot answer this, do not run the experiment. Every experiment supports at least one claim from templates/claims-experiments-matrix.md; anything else is a distraction you will not have time for.

Claim: — (none yet; this is the Phase 1 variance characterization, which supports the methodology section rather than a claim)

Purpose

Establish the Phase 1 baseline and characterize run-to-run variance.

This is the experiment that decides whether the rest of your thesis is measurable. Run the same benchmark 30 times unchanged and look at the distribution. If variance exceeds the effect size you hope to demonstrate, stop and fix the measurement. No amount of later analysis recovers from this, and it is the Phase 1 gate for exactly that reason.

What to record

• Median and IQR of the primary metric
• Coefficient of variation across runs
• The warm-up curve: metric over time for one long run, showing where it flattens
• Whether steady state holds across the measurement window

Result

Detectable effect size given this variance:

Verdict: (variance acceptable / must improve environmental control before proceeding)

Baseline tuning log

What you tuned, to what, and why. This becomes a paragraph in your methodology section and it is the answer to the question most likely to be asked about your evaluation.