LG AI 'EXAONE' Achieves World-Class Performance in Industrial Data Prediction, Surpassing Google and Salesforce

LG's artificial intelligence model, EXAONE, has demonstrated top-tier global performance in predicting industrial data, outshining major tech firms from the United States and China. This achievement highlights LG's technological prowess not only in general language models but also in industry-specific AI applications ready for immediate business deployment.

LG AI Research announced on August 7 that its EXAONE Tabular, designed for analyzing spreadsheet-like data, and EXAONE Forecast, which predicts future data trends, both secured the top spot (SOTA - State-Of-The-Art) on their respective global leaderboards. These international evaluation platforms use real-world industrial data to benchmark AI model prediction capabilities.

EXAONE Tabular excelled in the 'Categorical Data Prediction' category of the Tapprehna evaluation, achieving a score of 1760 points, surpassing Google's latest model at 1749 points. This model demonstrates high practical value by accurately detecting manufacturing defects or predicting financial risks with limited data. It can also fill in missing data points by analyzing surrounding information.

EXAONE Forecast also claimed the top position in a prediction evaluation led by Salesforce, outperforming Google and Alibaba in predicting data for unfamiliar fields without additional training. This capability allows for the analysis of complex, fluctuating indicators such as product demand, raw material prices, and stock prices using a single, versatile AI model.

"This case proves that accumulated data from various industrial sites, including manufacturing, can become a differentiated competitive edge for Korean AI," said Lim Woo-hyung, co-researcher at LG AI Research. LG intends to proceed with proof-of-concept (PoC) tests for these technologies in the manufacturing, bio-healthcare, and finance sectors starting in the second half of this year, with the goal of integrating them into actual business operations.

“

This case proves that accumulated data from various industrial sites, including manufacturing, can become a differentiated competitive edge for Korean AI.