Search papers, labs, and topics across Lattice.
Link枚ping University, Link枚ping, Sweden
3
3
4
12
Separating distractors into their own class leads to a significant 4.6 percentage point improvement in video object segmentation accuracy.
HyperSafe slashes harmful response rates in fine-tuned language models to below 1% without sacrificing task performance, revolutionizing safety alignment strategies.
Pre-trained vision encoders are shockingly bad at distinguishing identity when background context is controlled, but a simple contrastive learning scheme can fix it.
