PRX, the leading platform in AI research, in the fourth section of its observation series, disclosed its data strategy under the title "Nova". This strategy includes three main parts: collecting diverse data from public and proprietary sources, using self‑annotation techniques and pertaining to privacy, and finally, optimizing the processing pipeline to reduce repetition and increase access speed.
The result of implementing this program is a significant improvement in the accuracy of language and vision models; internal tests have shown that the variance in loss due to bruk usage will be reduced by up to 15%. The PRX team has announced that in the future it will integrate Nova with Iranian companies and international research centers to expand the repository of synthetic data.