Skip to content

feat(W43,LL'): INT2 activation witness W-106-G #923

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
gHashTag merged 1 commit into from
May 16, 2026
Merged

feat(W43,LL'): INT2 activation witness W-106-G #923

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 17 additions & 0 deletions assertions/int2_quant_witness.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,17 @@
{
"id": "W-106-G",
"wave": 43,
"freeze_date": "2027-01-15",
"claim": "INT2 activation variant of TTIHP27a degrades end-to-end accuracy by <= 2.0 pp on (MMLU + GSM8K + HellaSwag) averaged at identical weights, temperature 0.0, deterministic seeds, vs. INT4 activation baseline.",
"codebook": [-1.0, 0.0, 0.6180339887498949, 1.0],
"codebook_anchor": "phi^-1 cell of Sacred ROM",
"bio_mapping": "cortical-column-13 (occipital sparse hue gate)",
"reuses_opcodes": ["0xE8 SPARSE_SKIP", "0xED SPARSE_MASK", "0xE2 TOM"],
"new_l1_opcodes": [],
"new_l2_microcode": ["L2_COL13_INT2_GATE"],
"falsifier": "If 3-suite averaged accuracy delta > 2.0 pp on any official harness eval at freeze_date, W-106-G is REFUTED and W43 must be rolled back.",
"honest_lane": "R5",
"anchor": "phi^2 + phi^-2 = 3",
"doi": "10.5281/zenodo.19227877",
"silicon_vectors": ["S-177", "S-178", "S-179", "S-180", "S-181", "S-182", "S-183", "S-184"]
}
Loading