Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
1
0
3
2
Most vision-language models fail to correctly bind attributes to objects in complex scenes, but finetuning on synthetic text alone can fix this without harming general performance.