EXP-15 / completed / Paired system comparison

EXP-15: CM-7 versus CM-8 paired system comparison

69 CM-7 wins, 10 CM-8 wins and 21 draws. All 100 completed.

CM-7 White: 33W / 12D / 5L; CM-7 Black: 36W / 9D / 5L. Same frozen paired openings, BF16 decoding, adaptive helper budget and 256-ply cap; symmetric immediate draw claims. Both-player checkpoint and helper provenance is retained per game.

CM-7 supplied 137 helper overrides versus 19 for CM-8. These count decisions, not games. This comparison has no helper-free control.

These runs compare complete systems on selected fixed openings. They do not isolate input representation, supervision density, curriculum, initialization, inherited training, precision or helper/search configuration. They establish neither causal proof nor universal strength ranking. Unknown bounded searches are not proof of no mate.

All 100 PGNs · Audited player metrics · Comparison implementation · Replay.

Inspect the evidence

All run identities, checkpoint/suite hashes, protocols and results · Allowlisted source ZIP · Original/rendered source hashes · Training source · Evaluation source · Consolidated game viewer.

Original code is Apache-2.0; the external Rust helper retains its separate GPL terms. Host paths are redacted; datasets, optimizer state and private launch configuration are not bundled. Source identity is inspectable, but this is not a turnkey training replay. Missing game-start timestamps remain unknown.