From 752e113d67e510c4152bb1f33301b1dc509322af Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Tue, 28 Jul 2026 11:28:36 -0500 Subject: =?UTF-8?q?RESULT=2065:=20=E6=B3=84=E6=BC=8F=E4=BB=AA=E5=99=A8?= =?UTF-8?q?=E9=A6=96=E8=AF=BB+0.0685/6k(=E5=90=8C=E6=80=81=E5=90=8CLR,BP?= =?UTF-8?q?=E5=86=B7=E5=8A=A8=E9=87=8F=E4=BB=8D=E8=83=9C=3D=E4=BF=9D?= =?UTF-8?q?=E5=AE=88=E4=BD=8E=E4=BC=B0;=E8=BD=A8=E8=BF=B9=E7=A7=AF?= =?UTF-8?q?=E7=B4=AF=E6=85=A220=C3=97=3D=E5=89=8D=E8=BD=BD+=E8=87=AA?= =?UTF-8?q?=E6=84=88);=20=E7=AD=9B=E9=80=89=E7=94=B5=E6=B1=A01=E4=B8=89?= =?UTF-8?q?=E8=87=82=E5=9C=A8=E9=A3=9E(centered/K8/fp32)?= 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 | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'docs') diff --git a/docs/campaign/CASCADE_ABLATION_PLAN.md b/docs/campaign/CASCADE_ABLATION_PLAN.md index a73f28b..277ebf1 100644 --- a/docs/campaign/CASCADE_ABLATION_PLAN.md +++ b/docs/campaign/CASCADE_ABLATION_PLAN.md @@ -568,6 +568,18 @@ direction), not training-under-fault; wave-2 = co-training with faults injected +### RESULT 65 (2026-07-28): THE LEAK INSTRUMENT WORKS — first reading +0.0685/6k at the +SAME state (20x the trajectory-averaged rate: front-loaded/self-adapting), screening battery 1 +in flight. Paired 6k segments from fw135m_bsign_s250000: BP tail val_mean 3.4982 vs EP-plain +3.5668. Caveats favoring conservatism: BP arm ran COLD momentum (weights-only resume) yet +still outpaced by 0.0685 -> underestimate of BP's edge; the trajectory-integrated gap accrues +20x slower (EP re-equilibrates along its own path — the leak is partially self-healing, which +is WHY the run survives at all). As a SCREENING instrument: 1.5h/arm, single GPU, huge SNR. +BATTERY 1 (all from s250000 warm, 6k, vs the two bars): leak_cent (centered — if the exact- +symmetric read still leaks, the entire estimator-bias family is out); leak_k8 (deeper relax — +cos said saturated, the leak may live below cos resolution); leak_f32 (amp share on the same +instrument, cross-check of the ~12% window estimate). + ### RESULT 64 (2026-07-28): NOT A PLATEAU — A CONSTANT-RATE RELATIVE LEAK, 4-5x wider at C768; every named suspect cleared AT PROBE RESOLUTION; the leak-rate probe is the new instrument. LEDGERS CLOSED: 72M BP n=3: 3.2884/3.2943/3.2820 -> mean 3.2882+-0.006. Honest 72M gap: -- cgit v1.2.3