Search papers, labs, and topics across Lattice.
1
0
3
Small language models achieve high abstract reasoning scores by memorizing task distributions rather than learning generalizable rules, failing even simple out-of-distribution shifts.
