Search papers, labs, and topics across Lattice.
1
0
2
RL models not only outperform SFT counterparts in reasoning tasks but also develop a hierarchical architecture that enhances representational quality and processing efficiency.
