Search papers, labs, and topics across Lattice.
1
0
3
2
MADA-RL boosts compact model reasoning accuracy by 2% with 16 times fewer trainable parameters, redefining how critics learn from generators.
