Search papers, labs, and topics across Lattice.
Westlake University
2
0
6
12
Forget backprop and memory lookups: FAAST lets you adapt models at test time with a single forward pass, matching fine-tuning accuracy with massive speed and memory gains.
RLVR training leaves a tell-tale sign: prompts encountered during fine-tuning produce unusually similar reasoning trajectories, detectable without access to model internals.
