Ai2Labs
XYBench: Helping with the problem behind the question

XYBench is an 8,115-question benchmark testing whether LLMs spot misconceptions in user requests and redirect toward the user’s underlying goal.
Read at Ai2 ↗More from Ai2 on Accept All
LabsToyBlocks: A controlled world for robot learning and evaluation Ai2

LabsAi2 and Google collaborate on a new generation of climate models Ai2

LabsStateBench: Testing how language models track state changes over long documents Ai2

LabsHow Olmo 3 earned the highest ethical rating from Nextcloud Ai2

LabsHow researchers are adapting olmOCR 2 for historical and archaeological archives Ai2
