Upstage Presents 7 Papers at the ICML 2023 DMLR Workshop

AI startup Upstage presented a total of 7 papers at the Data-centric Machine Learning Research (DMLR) workshop of ICML 2023, achieving the highest number of research results among domestic companies.

This research included "DMOps: Data Management Operations and Recipes," which proposed an efficient AI training data construction process for enterprises, as well as papers addressing the cost and efficiency of data construction. In particular, through joint research with Professor Heuiseok Lim's research team at Korea University, Upstage achieved a wide range of research outcomes, including data-centric machine translation, commonsense reasoning, analysis of the negative effects of synthetic data, and "Explainable Error Benchmark," a methodology for speech recognition post-processing.

Upstage also presented papers at major AI conferences such as ICLR, CVPR, and ECCV last year, and was recognized for its technical capabilities by being the only domestic startup to rank among the top 5 in a survey by Imperial College London in the UK on papers at global top-tier AI conferences.