Search papers, labs, and topics across Lattice.
3
0
5
Coarse-to-fine hierarchical priors rescue vision models from out-of-domain failure in fine-grained symbol recognition, yet standard flat classifiers unexpectedly reclaim dominance the moment few-shot supervision is introduced.
Unlocking historical auction catalogs with VLMs could revolutionize provenance research by automating the extraction of structured metadata from variable formats.
Achieve Transformer-level handwritten text recognition accuracy with 1.6-1.9x faster inference and 38-42% less memory by ditching attention for a linear-time RetNet architecture.
