Noah Ziems
@NoahZiems
2.1K Following 3.9K Followers
Heres the latest on FelonyBench. OpenAI currently projected to have 300 felonies by the end of August with Anthropic only at 100.
If exponential growth continues, as it always does with language models, by this time next year we're projected to have more felonies than there are observable stars in the universe (~10^24)
Show more
New addition as of this morning: Verifiers v1 now supports GEPA, allowing you to get very fast and robust performance gains before you do any RL.
Fun fact:
In 1936, Einstein submitted a paper, got peer-reviewed for the first (and only) time in his career, called the reviewer's notes "erroneous," withdrew the paper from the journal, and published elsewhere
Show more
Our recent work on Pedagogical RL is out!
