From c47abf122a2074250ff565b41f3cb3d47bf9b512 Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Fri, 24 Jul 2026 05:16:20 -0500 Subject: =?UTF-8?q?RESULT=2059b:=20C768=E5=A4=A9=E8=8A=B1=E6=9D=BF?= =?UTF-8?q?=E5=AE=9E=E6=B5=8B8-12e-3@150k=E2=86=921e-3@345k=20=E2=80=94=20?= =?UTF-8?q?=E5=90=8E=E5=8D=8A=E7=A8=8B=E8=B6=85=E9=A1=B62-3=C3=97=E8=BF=90?= =?UTF-8?q?=E8=A1=8C(=E9=9D=9E=E6=93=A6=E8=BE=B9),bsign=E4=B8=8B=E5=A4=A9?= =?UTF-8?q?=E8=8A=B1=E6=9D=BF=E7=85=A7=E6=B2=89(=E9=94=90=E5=8C=96?= =?UTF-8?q?=E8=87=AA=E9=A9=B1,=E5=AE=BD=E5=BA=A6=E6=96=B0=E7=89=A9?= =?UTF-8?q?=E7=90=86);=20res>0.02=E5=91=8A=E8=AD=A6=E5=B8=B8=E6=95=B0?= =?UTF-8?q?=E4=B9=9F=E4=B8=8D=E8=B7=A8=E5=AE=BD=E5=BA=A6;=20end15=E6=8D=A2?= =?UTF-8?q?end075(=CE=B2=3D7.5e-4)=E9=87=8D=E6=94=BE?= 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 | 17 +++++++++++++++++ 1 file changed, 17 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 04bade5..a0c8efb 100644 --- a/docs/campaign/CASCADE_ABLATION_PLAN.md +++ b/docs/campaign/CASCADE_ABLATION_PLAN.md @@ -568,6 +568,23 @@ direction), not training-under-fault; wave-2 = co-training with faults injected +### RESULT 59b (2026-07-24): C768 CEILING MEASURED — the run spent its second half 2-3x +ABOVE the ceiling, and the ceiling sinks UNDER BSIGN at this width. R59's "graze" and "stability +transfers" both CORRECTED. +rho-grid @C768 (fine grid 0.5-20e-3, K=30): ceiling beta* = 8-12e-3 @s150000 (3e-3 healthy, +4x margin) -> ~1e-3 @s345000 (3e-3 = 3x OVER; 1.5e-3 diverges, resK 1.1e-2) -> 1-1.5e-3 +@s435000. So: (1) not a graze — sustained over-ceiling operation, fail-soft only because bsign +cut the bias feedback and guards were silent (plain_nf regime: no explosion, no learning); +(2) the ceiling sinks even with bsign at C768 — sharpening itself (sigma 569->661) drives it; +72M's "debiasing removes the sink" narrows to "removes the BIAS-driven component" — width adds +a second, bias-independent sinking term. NEW LADDER PHYSICS: ceiling scale drops with width +(~1/sigma^2); at 270M+ even 1e-3 may cross mid-run -> the recipe needs per-stage beta +descent gated by measured ceilings (the ratchet/descent axis returns, this time with the +instrument to calibrate it). Also non-transferring: the res>0.02 legality constant (C768 ran +half its schedule at res~1e-2, under the alarm line, while the harvest died). +end15 (beta 1.5e-3) KILLED pre-data (measured above-ceiling at the resume state); +fw135m_end075 relaunched: s345000, beta=7.5e-4 (below both late readings), final 95k replay. + ### RESULT 59 (2026-07-24): FW135M COMPLETED 3.3839 — stability TRANSFERS (zero events, 440k, bsign), but the OPERATING POINT does not: beta=3e-3 grazed the C768 ceiling through the endgame and the anneal harvest was lost. Scaling verdict DEFERRED pending the beta-fix replay. -- cgit v1.2.3