Wilson Wu

An­nounc­ing the ARC White-Box Es­ti­ma­tion Challenge

Mechanis­tic es­ti­ma­tion for ex­pec­ta­tions of ran­dom products

ARC progress up­date: Com­pet­ing with sampling

• Wilson Wu 16 Feb 2025 17:32 UTC

  2 points

  0

  in reply to: Daniel Murfet’s comment on: Am­bigu­ous out-of-dis­tri­bu­tion gen­er­al­iza­tion on an al­gorith­mic task

  For that earlier section, we used smaller models trained on $S_4$ intersect $A_4\times 2$ (4,000 parameters) instead of $S_5$ intersect $A_5\times 2$ (80,000 parameters) -- the only reason for this was to allow for a larger sample size of 10,000 models with our compute budget. All subsequent sections use the $S_5$ models.

Am­bigu­ous out-of-dis­tri­bu­tion gen­er­al­iza­tion on an al­gorith­mic task

The sling­shot helps with learning