Search papers, labs, and topics across Lattice.
Stanford University
1
0
2
Models that write intermediate states significantly outperform those that only report final answers, achieving up to 91% accuracy in predicting outcomes from edited states.
