Issue #97 producers: Phase A calibration (57dfcb7) + Phase B holdout-only authority (c60a4b7) — terminal evidence retained - #31
Merged
ezutfen merged 4 commits intoSep 6, 2026
Conversation
…dmission entrypoints) Adds the separate, explicitly authorized h95 holdout path promised by the frozen Phase A producer's admission docstring (commit ce7eedf): benchmarks/inferswarm_97b/ holdout-only entrypoints accept exactly the 24 committed h95 cells and delegate to the byte-identical frozen Phase A reference/chain runners with ONLY the admission callable swapped. - Phase A calibration producer 57dfcb7... unchanged; every frozen inferswarm_97 file byte-identical (enforced by test). - No methodology, threshold, topology, semantic, or evidence-contract change; no execution/model math duplicated (source-audited by test). - 13-test CPU-only suite (synthetic fixtures only, no secret material). - Phase A harness suite (10 tests) still green.
ezutfen
changed the title
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Two frozen producers for the accepted terminal FlashML-org#97 campaign
These exact producer commits generated the accepted terminal evidence for InferSwarm issue FlashML-org#97 (
V4_HOLDOUT_CORE_FAIL). This PR should now be merged to retain the exact producer implementation history.Phase A calibration producer —
57dfcb7289efac8f66de5b3abbe8de04f2580f75(immutable)The c95/p95 physical campaign harness (
benchmarks/inferswarm_97/): RTX 3090 reference runner, teacher-forced three-stage RTX 3060 chain runner, last-stage service. Generated all Phase A calibration evidence and the frozen core limits/telemetry bands. Remains byte-identical since the campaign freeze.Phase B holdout-only producer —
c60a4b7080913eba90585de8e8602460a7f5b1a7(one additive commit on top)Adds
benchmarks/inferswarm_97b/: holdout-only entrypoints that admit exactly the 24 committed h95 cells and delegate to the frozen Phase A runnermain()with only the admission callable swapped, plus a 13-test CPU-only equivalence suite (Phase A byte-identity, admission matrix, delegation semantics, no-execution-math source audit; synthetic fixtures only). No execution/model math changed — this producer executed the accepted h95 holdout (reference 24/24, candidate 24/24 teacher-forced).State
inferswarm-research; both producer SHAs are externally pinned in InferSwarm's retained execution-authority records.docs/qualification/gemma4-12b-it-v4-campaign-97/.