TLDR the checklist? Now you can just watch it.
Big shoutout to @macketing_ai for making this! I believe this is also the first animated version of our mascot.
The agentic checklist runs in an automated loop via two agents: A Discovery Agent that investigates past AutoScientist experiments and generates modular requirements based on critiques, and a Verification Agent that applies a tailored checklist to each and every training example.
Show more
Much of the datasets we work with at @adaption_ai are from domains that are not verifiable. As you can imagine, this makes data eval difficult given there are no ready-made verifiers available like in math and code! This is the problem we’re tackling with the agentic checklist:
Show more
