Search papers, labs, and topics across Lattice.
2
0
4
Frontier-scale agentic RL on 700B+ MoEs is no longer locked inside hyperscaler proprietary stacks, achieving stable 263-second step times across 64 GB300 GPUs via a verified, open-source training infrastructure.
Bridging the gap between inference and adaptation in VLMs could lead to significant performance boosts by ensuring robust pseudo-labels that accurately reflect sample-level relationships.
