igneum-network/igneum
1
0
Fork 0
Code Issues Pull requests Releases Activity

class v5 design page: section 14 carries adv-accept's multiplicity reading at the acceptance sample

Browse source 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
igneum-labs 2026-10-07 19:40:48 +00:00
parent a64aafb488
commit 80736ef56b
1 changed files with 2 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
docs/design/class-v5-stored-state.md
View file

@ -297,4 +297,6 @@ The exemplar under class v5 itself (adv-accept, 20:36 UK, logs/adv-accept/v5-exe
The known-failed test's reading on box 2 (20:33 UK): class v4 accepts the exemplar with site 6 at 0.9919 (1,031,917 distinct of 1,048,576, adv-accept's exact count; index 0 the most read, 235 times at the 2^20 random sample against 1,677 per 10^6 sequential nonces); class v5 refuses it at attempt 2 naming site 6 at 0.991; the class v5 draw lands on attempt 4 (program 734fbb8e3e4cd20f) and passes; the class v4 draw still lands on attempt 2; the genesis draws of both classes read 0.9990 at their minimum site.
The multiplicity reading at the acceptance sample (adv-accept's gap tool, 20:4x UK, logs/adv-accept/gap-rep8.log): the exemplar's site 6 spends 0.168 percent (closed form) and 0.151 percent (live day) of its 2^20 reads on word indices read 8 or more times, every other site 0.000 percent; seed 100211 (d07885a437e237c7) reads site 9 at 0.186 / 0.179 percent with ratio 0.9907 / 0.9908; the control program reads 0.000 at every site. Items are 16 words and 2^24 of them, so 3.35 percent of a site's reads into the top 0.1 percent of items at 2^24 sequential nonces reads as 0.17 percent on heavily repeated word indices at 2^20, and the distinct count the floor reads carries the same signal with the model's spread of 0.0001 under it. The multiplicity statistic would be the second instrument if the floor's clean rejection rate read high; the census decides.
What the floor does not reach, named: a hot set whose excess collisions at 2^20 stay under 0.5 percent of a site's evaluations (a top-0.1-percent item share under about 1.3x); adv-accept's lowest-ratio list between 0.995 and 0.998 reads 1.29x to 1.63x live on the f8 gate but no hot set by X_f >= f, and the floor rejects those too (cheap: one more attempt). A chip that holds 0.1 percent of a window's items serves at most the excess the floor allows, 0.5 percent of a site's reads, 1/16 of that over the program: under the 1.067x bound of AP-F8-1 by a wide margin. The per-site hot-item test at live scale stays owed only if the census reads the floor's clean rejection rate over 1 percent.