fix(golden): precision-scoped parity classification + scoped claims - #211
Merged
Merged
Conversation
…pe the parity claims
The golden test classified models by precision substring in the GOLDEN
FILENAME - but tha-g2p-small-1.0 resolves to an int8 zip whose golden
name drops the suffix, so a quantized artifact got byte-exact
assertions and failed on a near-tie flip ('taj' vs 'ta:j' class).
Classification now reads the loaded manifest's precision field
(fp32/fp16 exact; int8/int4 decode-health length floor). Full
GOLDEN_DIR run: 12 models x 25 rows, green.
Paper A, paper C, and publication notes now state the parity
guarantee precision-scoped: byte-identical for fp32/fp16 on all
hardware tested; measured quality parity for quantized artifacts.
RESULTS.md gains the speculative-decode acceptance entry (0.9886
mean, 8.85 tokens/verifier pass, output-preserving).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes the golden-v1 workstream (rababa TODO.impl/08) and scopes every parity claim to what was measured.
Test fix. test_golden_matrix classified artifacts by a precision substring in the golden FILENAME. tha-g2p-small-1.0 resolves to an int8 zip whose golden name drops the suffix - a quantized artifact under byte-exact assertions, failing on a textbook near-tie flip (one haraqat byte). Classification now reads the loaded artifact's manifest precision: fp32/fp16 get byte-exact assertions; int8/int4 get the decode-health length floor from the measured cross-hardware spread. Full local GOLDEN_DIR run green: 12 models x 25 rows.
Claim scoping. Paper A (5 sites), paper C, and PUBLICATION-NOTES now say byte-identical for full-precision artifacts and measured quality parity for quantized ones - matching How-we-measure and the 2026-09-07 RESULTS entry. The golden-v1 release body was annotated with the same scoping.
RESULTS.md gains the speculative-decode acceptance entry behind the playground tier shipped yesterday: 25/25 golden rows, acceptance 0.9886 mean (min 0.955), 8.85 tokens per verifier pass at K=8, output-preservation verified byte-identical on all 18 exactness-checked rows.