Search papers, labs, and topics across Lattice.
1
0
3
Freezing a VLM's vision encoder and adapting it for end-to-end driving surprisingly beats fine-tuning, suggesting pre-trained general knowledge trumps domain-specific adaptation for robustness.
