Search papers, labs, and topics across Lattice.
3
0
7
12
Even the best vision-language models struggle with reliable evaluation of computer-using agents, but OS-Shepherd models offer a low-cost solution that matches their performance.
Book-level organization of synthetic data boosts language model performance by over 1% compared to traditional methods, highlighting the power of structured content.
OpenMobile proves that high-performing mobile agents can be trained on entirely synthetic, open-source data, closing the gap with closed-source models and enabling broader research.
