MedGaugeArticles
Shorter pieces on why MedGauge measures the way it does. Each carries an evidence grade and names the methodology version it follows; hypotheses are labelled as such and will be revised when results are available.
| Article | Kind | Evidence | Date |
|---|---|---|---|
| Binary pass, continuous score, and how rows are ordered Why MedGauge reports a binary pass and a continuous score for every run, orders rows by pass^3, and groups them by paired comparisons rather than by overlapping intervals. | Method | Analysis | 2026-10-29 |
| Why a program, not a judge model MedGauge grades end states with deterministic verifiers. This article explains what that buys, what it costs, and where it stops. | Method | Analysis | 2026-10-29 |
| Answer-key quality comes first Label errors in the test sets of widely used benchmarks are pervasive and destabilise their rankings. MedGauge treats the key as the first thing to verify, and measures how often keys change. | Note | Verified | 2026-10-29 |
| The harness is part of the measurement Why every MedGauge result names a harness, and what we expect to learn by measuring the same model under different harnesses. | Reasoning | Hypothesis | 2026-10-29 |
| When guidelines disagree How grading references handle a clause on which sources from different jurisdictions point in different directions, without marking a correct answer wrong. | Method | Analysis | 2026-10-29 |
| Why a corrected value is an automatic fail In audit and quality-control roles, writing the "right" value into the record destroys the evidence the review exists to produce. MedGauge treats it as a veto. | Method | Analysis | 2026-10-29 |
| Predict before you test Writing down expected pass rates per model tier before a task is run, and publishing the difference afterwards, keeps difficulty claims honest. | Reasoning | Hypothesis | 2026-10-29 |
What the evidence grades mean
- Verified
- Every statement the article makes about external work cites a source a reader can open (a persistent identifier or a stable public page), and each such statement has been checked against the source's text or publisher abstract on the date shown on the article. Every statement about MedGauge's method matches the methodology version the article names.
- Analysis
- The article states MedGauge's rules and the reasoning for them, or reports MedGauge's own computations. It makes no unsourced claim about external work, and it is not evidence about any model or harness.
- Hypothesis
- A statement to be tested that has not yet been tested against MedGauge results.
An article that no longer meets its grade is regraded, not left as it stands.