Control
First draft
Focuses the comparison on live construction choices instead of later cleanup or editorial smoothing.
Risk
Shortcuts show
Makes repetition, padding, and easy structural moves harder to hide inside the totals.
Evidence
Robustness
Asks whether a ranking survives clear, repeatable constraints rather than one perfect setup.
Evidence Frame
Core conditions
The Solo Dataset is framed as first-draft, unaided, non-repeating work produced under pressure.
Those conditions are not decoration. They define the test environment:
- solo-authored
- first-draft baseline
- no AI-written lyrics
- no dictionary dependence
- no thesaurus dependence
- no rhyming dictionary
- non-repetition pressure
- no intentional recycled lines
- produced under time compression
What the conditions control
The constraints keep the public findings from being read as edited anthology results, assisted composition results, or cherry-picked quotation results.
They also explain why repetition control matters. In a first-draft, non-repeating environment, recycled lines, filler structures, and narrow vocabulary loops are not harmless quirks. They are shortcuts the metrics are supposed to catch.
Why this matters
The constraints do not excuse the result. They raise the difficulty of the result.
A strong score under loose conditions is one thing. A strong score under first-draft, unaided, non-repeating pressure is a different claim. This page explains the pressure so the evidence can be read in the right frame.
What the conditions do not prove
The constraints do not ask the reader to accept private claims on faith. Public pages still rely on aggregate charts, public-safe labels, validation checks, and controlled review paths.
The condition set explains the test environment. The charts show whether measured behavior survives that environment across metric families, slice sizes, tokenizer lenses, and exception checks.
How to read the charts
Start with the coverage dashboard to understand the measurement surface: workbooks, metrics, tokenizer families, segments, and slice sizes.
Then use tokenizer robustness and cross-slice stability as fairness checks. A result is stronger when it stays visible under different tokenizers and does not depend on one convenient slice.
Exception count shows where The Sequel is not first, which keeps the constraint frame honest. The interaction-score leaderboards then show how the headline index family behaves when the conditions and metric families are considered together.
Public-safe limits
The public constraints page describes writing conditions and aggregate verification behavior. It does not publish raw writing, protected excerpts, private manifests, source-level mappings, third-party source labels, artist names, album titles, or song titles.
Public-safe boundary
Public pages show aggregate evidence, metric behavior, method provenance, and corpus structure. Protected text, identities, source titles, and reconstructable mappings stay private.