Super proud to have contributed to this amazing team effort, led by @sreejan.bsky.social and @botoscsabi.bsky.social!
We asked: Which AI models learn to play video games like humans, comparing both behavior and internal representations. The answer surprised us!
Check out our paper and post below
Marcelo Mattar
Jeee 🐦⬛
I am very proud of our joint effort with @sreejan.bsky.social on the project "Reason to Play"
LRMs show human-like rule discovery, and their hidden states predict human brain activity during gameplay 10x better than previous methods
Interactive demo + paper:
botcs.github.io/reason-to-pl...
32 fMRI-scanned humans and 8 frontier open weight LLMs play ARC-AGI like games with no rules given. The reasoning models match the human learning trajectories and their hidden states predict human bra...
