✳flâneur — a map of the web's best reading

Getting 50% (SoTA) on ARC-AGI with GPT-4o — LessWrong

lesswrong.com · 11,643 words · saved by 1 readers

I recently got to 50%[1] accuracy on the public test set for ARC-AGI by having GPT-4o generate a huge number of Python implementations of the transfo…

x Getting 50% (SoTA) on ARC-AGI with GPT-4o — LessWrong AI Capabilities GPT AI Frontpage 267 Getting 50% (SoTA) on ARC-AGI with GPT-4o by ryan_greenblatt 17th Jun 2024 16 min read 50 267 I recently got to 50% [1] accuracy on the public test set for ARC-AGI by having GPT-4o generate a huge number of Python implementations of the transformation rule (around 8,000 per problem) and then selecting among these implementations based on correctness of the Python programs on the examples (if this is confusing, go here ) [2] . I use a variety of additional approaches and tweaks which overall substantial

Explore this link on the map →

related reading