Search papers, labs, and topics across Lattice.
Affiliation:
2
0
3
0
HALO achieves a 13.7 percentage point boost in zero-shot open-set accuracy with 10x fewer parameters than leading models, redefining efficiency in human activity recognition.
Flow-matching models can now be fine-tuned with GRPO using only a third of the training steps, thanks to a novel off-policy approach.
