Search papers, labs, and topics across Lattice.
3
0
8
0
RL-driven adaptive batching can yield a staggering 348% throughput improvement in multi-GPU settings, far surpassing traditional heuristics.
Action-conditioned predictions from a compact latent model reveal that diffusion methods can dramatically outperform traditional regression in scene forecasting for autonomous vehicles.
Forget fixed retrieval: Offline RL agents can learn to selectively retrieve policy information, slashing retrieval costs by nearly 50% while maintaining state-of-the-art decision accuracy in prior authorization.
