Evidence Context
The finding is supported across aggregate leaderboards, slice-size gap behavior, separate and combined metric comparisons, pairwise results, and first-place rates.
Importance of scale
Short excerpts can produce unstable metric peaks or overstate a local pattern. The Sequel uses an entire half-corpus, allowing the result to be tested across many slices and tokenizers instead of depending on a favorable passage or isolated window.
Late-Stage Separation
The Sequel demonstrates that PRM’s late-stage material does not merely preserve the earlier baseline. It separates from it across the tested metric family, showing measurable growth in complexity, lexical discipline, and repetition control as the solo-authored corpus expands.
Finding Scope
Across the displayed metric family, PRM Half 2 leads the tested comparison pool under shared preprocessing, tokenizer, slice, and aggregation conditions.
The finding applies to the analytical framework reported here, while protected source text and private mappings remain within controlled review.
Reference Comparison Boundary
Reference corpora provide fixed analytical coordinates for interpreting PRM under shared preprocessing, tokenizer, slice, and metric conditions. Public charts identify public-domain references where permitted, while protected comparator identities and source mappings remain within controlled review.