From 426df3f015868a1f782ed86711afc18108992068 Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Fri, 24 Jul 2026 05:11:42 -0500 Subject: =?UTF-8?q?RESULT=2059:=20135M=E5=AE=8C=E8=B5=9B3.3839=20=E2=80=94?= =?UTF-8?q?=20=E7=A8=B3=E5=AE=9A=E6=80=A7=E8=BF=81=E7=A7=BB(=E9=9B=B6?= =?UTF-8?q?=E4=BA=8B=E4=BB=B6440k)=E4=BD=86=E5=B7=A5=E4=BD=9C=E7=82=B9?= =?UTF-8?q?=E4=B8=8D=E8=BF=81=E7=A7=BB:=CE=B2=3D3e-3=E6=94=B6=E5=AE=98?= =?UTF-8?q?=E6=93=A6C768=E5=A4=A9=E8=8A=B1=E6=9D=BF(=CF=83661,cos=E9=80=80?= =?UTF-8?q?=E5=88=B00.968,=E6=9C=80=E5=90=8E93k=E9=80=80=E7=81=AB=E9=A2=97?= =?UTF-8?q?=E7=B2=92=E6=97=A0=E6=94=B6,best@79%);=20scaling=E5=88=A4?= =?UTF-8?q?=E5=86=B3=E6=8A=BC=E5=90=8E;=20=E6=95=99=E8=AE=AD=3D=E7=AA=97?= =?UTF-8?q?=E5=8F=A3=E5=B8=B8=E6=95=B0=E4=B8=8D=E8=B7=A8=E5=AE=BD=E5=BA=A6?= =?UTF-8?q?,=E6=AF=8F=E7=BA=A7=E9=98=B6=E6=A2=AF=E9=9C=80=CE=B2=20re-gate;?= =?UTF-8?q?=20=CF=81=E6=8E=A2=E9=92=88+=CE=B21.5e-3=E6=94=B6=E5=AE=98?= =?UTF-8?q?=E9=87=8D=E6=94=BE=E5=8F=8C=E7=AE=A1=E9=BD=90=E4=B8=8B?= 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 | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) diff --git a/docs/campaign/CASCADE_ABLATION_PLAN.md b/docs/campaign/CASCADE_ABLATION_PLAN.md index 55b6977..04bade5 100644 --- a/docs/campaign/CASCADE_ABLATION_PLAN.md +++ b/docs/campaign/CASCADE_ABLATION_PLAN.md @@ -568,6 +568,28 @@ direction), not training-under-fault; wave-2 = co-training with faults injected +### 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. +Numbers: best 3.3839 @347.2k (79% of schedule; final 93k deep-anneal steps bought NOTHING; +val_mean flat 3.593->3.581 over the last 300k). Grazing signature, monotone by window: +cos 0.990->0.981->0.968, drift 0.0097->0.0173, gn 0.146->0.197, while sigma climbed to 661 +(72M endgame: 475, cos 0.995). Mechanism: wider model -> larger sigma -> ceiling ~1/sigma^2 +LOWER, and the 72M-calibrated beta=3e-3 met it. bsign held (zero d-skips, no storm, no +collapse) = the stability claim extends to 135M; but estimator quality degraded in the graze +and the last-fifth harvest vanished. Endpoint sits INSIDE the 72M seed band (bsign mean +3.364+-0.03) = a FLAT scaling step as measured, n=1, attribution impossible without the +135M BP twin (collaborator, now urgent). +LESSON (ladder doctrine): window constants do NOT transfer across width — every rung needs a +beta re-gate (rho-grid at that width), exactly as BASELINE_SPEC prescribed for depth events. +Candidate width rule: beta ~ 1/sigma_end^2 (C768 sigma_end/C512 ratio^2 = 1.94 -> beta ~1.5e-3). +IN FLIGHT: (a) probe_cx3_rhogrid on fw135m ckpts s150000/s345000/s435000, fine beta grid — +the direct ceiling measurement at C768 (physics certificate for the graze); (b) fw135m_end15: +resume s345000 (just before best), beta 1.5e-3, replay the final 95k (2-GPU, ~16h). Harvest +returns (best << 3.3839) -> diagnosis confirmed + better crown + the width-scaling beta rule +enters the recipe; no return -> C768 endgame is genuinely harder, escalate to estimator +(centered tail) or schedule work. + ### RESULT 58 (2026-07-22): BSIGN SEED-2 — STABILITY REPLICATED (n=2, zero events both runs); CE mean 3.3637 +- 0.029, s1 was the lucky draw (dip-statistics bite #4). fw72m_bsign_s2 DONE 3.3924: full 234k, d0/g0/r0 end to end under the res-alone gate — the -- cgit v1.2.3