From 07e06375620669e1fb3d36ded64a5867645e423d Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Sat, 18 Jul 2026 22:47:19 -0500 Subject: =?UTF-8?q?RESULT=2048b:=20=E4=B8=80=E8=88=AC=E5=8C=96BBP(amp+?= =?UTF-8?q?=E6=B5=8B=E8=B0=B1,=E5=85=8D=E7=BB=93=E6=9E=84=E5=81=87?= =?UTF-8?q?=E8=AE=BE)=3DR>1=E5=A4=84=E5=A4=84=E6=88=90=E7=AB=8B=E5=88=B01e?= =?UTF-8?q?-4,EP=E9=99=84=E5=8A=A0=E5=99=AA=E5=A3=B0=E4=BD=8E=E4=BA=8E?= =?UTF-8?q?=E4=BF=A1=E5=8F=B75-50=C3=97;=E6=95=B0=E5=AD=97=E5=9F=9F?= =?UTF-8?q?=E4=B8=8B=E7=95=8C=E9=9D=9E=E6=A3=80=E6=B5=8B=E7=9B=B8=E5=8F=98?= =?UTF-8?q?(bf16=3D=E4=BD=8D=E7=A7=BB=E8=88=8D=E5=85=A5=E6=82=AC=E5=B4=96?= =?UTF-8?q?=E9=9D=9E=E5=99=AA=E5=A3=B0bulk);=E5=8D=95=E6=AD=A5=E6=A3=80?= =?UTF-8?q?=E6=B5=8B=E7=93=B6=E9=A2=88=3Dbatch=E5=99=AA=E5=A3=B0(=E4=B8=8E?= =?UTF-8?q?BP=E5=85=B1=E4=BA=AB);=E7=9C=9F=E7=9B=B8=E5=8F=98=E5=9C=A8?= =?UTF-8?q?=E6=A8=A1=E6=8B=9F=E8=AF=BB=E5=87=BA/=E9=87=8F=E5=8C=96?= =?UTF-8?q?=E5=9F=9F,qb6=20=CE=B2-buyback=E5=8D=B3=E7=A9=BF=E8=BE=B9?= =?UTF-8?q?=E8=AF=81=E6=8D=AE?= 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') diff --git a/docs/campaign/CASCADE_ABLATION_PLAN.md b/docs/campaign/CASCADE_ABLATION_PLAN.md index c74a8dd..399d8d6 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 48b (2026-07-18): GENERALIZED BBP (amp semantics, measured-spectrum, no iid +assumption; user-directed): R = spike/edge > 1 EVERYWHERE down to beta=1e-4 — EP's extra noise +sits 5-50x BELOW the signal spike in bf16-amp too. The digital floor is NOT a detectability +transition; the binding single-step detection limit is BATCH noise (shared with BP). +probe_bbp2 @s150000, 8 layers x 8 batches x beta {1e-4..1e-2}: R@1e-4 = 8.7-46.8 per layer, +FLAT toward low beta (no 1/beta bulk: our theta-read never differences energies numerically — +beta enters via state displacement, so bf16 bites as an all-or-nothing ROUNDING CLIFF at +beta*||d|| ~ ulp(z) [RESULT 11's death], not as a noise bulk). u1-overlap ~0.37 (b0-b8) / +0.66-0.80 (b11) beta-INDEPENDENT => dominated by batch sampling noise (missing control column: +per-batch BP u1 vs mean-BP u1 — same-batch cos(EP,BP)~0.99 implies EP u1 ~ BP u1, so 0.37 is +the batch-noise number, not an EP defect). WHERE the sharp BBP transition genuinely lives: +(a) ANALOG readout (ENOB additive bulk -> T64 design equation stands), (b) aggressive weight/ +compute QUANTIZATION — qb6 beta-buyback is an in-hand crossing of the edge; controlled demo = +rerun bbp2 with 6-bit qcomp, expect R to cross 1 and overlap to jump with beta. (c) bf16 cliff +as the degenerate case. beta_floor's empirical benefit now fully attributed to window geometry +(distance from cliff + T2-sensitivity), NOT detectability. + ### RESULT 48 (2026-07-18): BBP FLOOR AUDIT — the fp32 SIMULATOR has NO additive detectability floor (a=0 all layers, top-direction overlap 1.000 down to beta=3e-4); BBP becomes a HARDWARE design equation. probe_bbp @plain s150000, 24 layers (qkv+w2 x12), 6 batches x beta {3e-4,1e-3,3e-3,1e-2}: additive coefficient a = 0 within fit resolution EVERYWHERE; residual error is beta-independent -- cgit v1.2.3