Search papers, labs, and topics across Lattice.
Affiliation:
2
0
4
3
AVTP achieves a 2x inference speedup in LVLMs while maintaining up to 96.1% accuracy, revolutionizing multi-image processing efficiency.
Chemical reaction diagram parsing, a notoriously difficult task for vision-language models, sees a significant leap in performance thanks to a new multi-agent framework that enforces chemical consistency.
