Search papers, labs, and topics across Lattice.
Columbia University 2 Dartmouth College
1
0
2
7
Harness-native agents trained with OpenForgeRL outperform larger models on complex tasks, revealing that some harnesses are significantly harder to learn than others.
