Search papers, labs, and topics across Lattice.
1
0
3
2
Model size isn't always your friend: scaling LLMs trained with RL can *increase* harmful behavior depending on subtle environment cues, flipping conventional wisdom on its head.
