Search papers, labs, and topics across Lattice.
JD.com
3
0
6
Transforming off-policy tokens into on-policy ones could be the key to unlocking more robust and efficient alignment for large language models.
FlowTrain redefines VLM training efficiency, achieving up to 1.7x throughput improvements by decoupling execution and optimizing resource allocation.
Training trillion-parameter recommendation models at scale doesn't have to be bottlenecked by data movement: NestPipe achieves 3x speedup on 1500+ accelerators by overlapping communication and computation.
