o3 varietal beaten by an o3 varietal for top spots on our benchmark

@jam3scampbell @aidan_mclau it’s from a scene in three body problem

@aidan_mclau @teortaxesTex 😭💗 omg what

@aidan_mclau @teortaxesTex immense approval of this only bc I fall into that

@growing_daniel effiel

@hallerite @jam3scampbell @max77sabers we very simply don’t have the funds for this 😄😃😀

@jam3scampbell James hahahahaha

@SpencerKSchiff 😎

@MillionInt wait which part 😅 like why o3 scored so high?

@buildanything were you joking or serious bc what do you mean trust me with this important work, I co built the important work 😭

@_xjdr thank you 🥹 + yes

@Mrcfyz @aidan_mclau @jam3scampbell the big reason why models “fail” on our benchmark is because of a low novelty score / a duplicate answer a model is returning as a result the coherence judge lowkey does nothing planning to add a mix of judge models in future for the lab bias issue tho

@aidan_mclau three more you may or may not know: I miss you

@buildanything @aidan_mclau @jam3scampbell girl wtf leave me alone 😭

note: we developed the methodology long ago and left it largely the same // aidanbench was never impacted by where Aidan was working but since it’s called…yk…Aidan-Bench + o3 leading rn, we could smell the “you rigged it” from miles away high integrity decision by Aidan

@SpencerKSchiff hmmmmm

@AURORAmusic @Spotify Conquerer next , you divined a masterpiece with that 😭💗

@SpencerKSchiff @multimodalagent you guys are foul his real name is perfect as is 😭😭✋🏽

@SpencerKSchiff @multimodalagent No it doesn’t it’s just shorter 😭 McLaughlin erasure is not appreciated by me it literally has laugh in it