Search papers, labs, and topics across Lattice.
1
0
3
VLMs reason better when shown visually similar examples with synonymous questions, enabling a new self-training approach that beats state-of-the-art visual reasoning datasets.
