From 428f5b4d0d7c3068e64b60903e101e82a5a2fafb Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Mon, 27 Jul 2026 00:59:43 -0500 Subject: =?UTF-8?q?RESULT=2062:=20C768=E9=94=9A=E7=82=B9BP=3D3.0902,EP=20g?= =?UTF-8?q?ap=E7=AC=AC=E4=B8=80=E7=BA=A7=E9=98=B6=E6=A2=AF=E7=88=86?= =?UTF-8?q?=E7=82=B86=C3=97(0.047=E2=86=920.294);=20BP=E5=81=A5=E5=BA=B7?= =?UTF-8?q?=E8=B5=9A=E6=BB=A1-0.198=3D=E9=85=8D=E6=96=B9/=E6=95=B0?= =?UTF-8?q?=E6=8D=AE/=E5=BD=A2=E7=8A=B6=E6=97=A0=E7=BD=AA,=E5=AE=BD?= =?UTF-8?q?=E5=BA=A6=E9=97=AE=E9=A2=98100%EP=E7=89=B9=E6=9C=89;=20caveat?= =?UTF-8?q?=3D=E5=B9=B2=E5=87=80C768=20EP=20run=E5=B0=9A=E4=B8=8D=E5=AD=98?= =?UTF-8?q?=E5=9C=A8;=20=E5=AE=9A=E4=BD=8D=E6=8E=A2=E9=92=88=E4=B8=8AGPU0?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn --- docs/campaign/CASCADE_ABLATION_PLAN.md | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'docs/campaign/CASCADE_ABLATION_PLAN.md') diff --git a/docs/campaign/CASCADE_ABLATION_PLAN.md b/docs/campaign/CASCADE_ABLATION_PLAN.md index 426410a..d467cb0 100644 --- a/docs/campaign/CASCADE_ABLATION_PLAN.md +++ b/docs/campaign/CASCADE_ABLATION_PLAN.md @@ -568,6 +568,19 @@ direction), not training-under-fault; wave-2 = co-training with faults injected +### RESULT 62 (2026-07-27): THE C768 ANCHOR — BP 3.0902; the EP gap EXPLODES 6x at the +first ladder rung (0.047 -> 0.294, ~34% ppl). Width problem is 100% EP-specific; recipe/data/ +shape exonerated (BP gains the full expected -0.198 from 72M). +fw135m_bp (exact EP twin, 440k/2.7B tok, seed 1): 3.0902. Ledger: 72M BP 3.2884/3.2943 (n=2, +band 0.006) | 135M EP best-of-3-arms 3.3839 (3e-3 lineage; 1.5e-3 rescue 3.4643; 7.5e-4 +endgame 3.4805). HONEST CAVEAT: no clean C768 EP run exists — first run spent its second half +over a sunken ceiling, b15 was a mid-run rescue with constant backstop halvings and an unknown +limiter; "current recipes fail at width" is measured, "EP fails at width" is NOT yet. The +scaling thesis now hangs on whether a width-correct recipe exists inside the C768 corridor. +NEXT (GPU0, user holds 1/3): gap DECOMPOSITION at C768 — probe_blockcos on fw135m_bsign +s150000 (healthy phase): if the gap still lives in top-block transmission but scaled up, the +width amplification of the known transmission bias is the target; then 72m_bp_s3 clean restart. + ### RESULT 61b (2026-07-25): CORRECTION — the sentinel's 21 DIVERGE alerts were FLOOR-NOISE FALSE POSITIVES; b15's corridor at 1.5e-3 was HEALTHY through 300k+; its true limiter is UNKNOWN. Row audit: s255000 @1.5e-3 resK=1.40e-6 (= fp floor, fully converged), rho_tail 1.0032 — the -- cgit v1.2.3