LG AI Research said Friday its Exaone models for tabular data analysis and time-series forecasting had topped global AI benchmarks, outperforming competing models from tech giants including Google and Alibaba.

The benchmarks measure how well AI models make predictions using data from industries such as chemicals, healthcare, finance and manufacturing.

Exaone Tabular, LG’s foundation model for structured data, posted an ELO rating of 1,760, ahead of Google’s TabFM at 1,749.

Trained on large volumes of synthetic tabular data, the model is designed to directly understand relationships among rows, columns and data points. General-purpose language models, by contrast, typically convert tables into text before analyzing them.

LG’s time-series model, Exaone Forecast, meanwhile ranked first in zero-shot forecasting, outperforming models from Google, Alibaba and other global technology companies, according to the institute.

Zero-shot forecasting measures whether an AI model can accurately predict data from an unfamiliar field without additional training.

Exaone Forecast can handle time-series data across finance, energy, manufacturing and healthcare with a single model, reducing the need to develop separate forecasting systems for individual industries or use cases.

LG AI Research is already using the model across LG affiliates to forecast product demand at LG Electronics and raw material prices at LG Energy Solution.

The institute has also been working with Koscom and the London Stock Exchange Group since this year to provide AI-based analysis of the South Korean and US stock markets.

“The focus of global tech companies is rapidly shifting from general-purpose language models toward industry-specific AI that can solve problems in actual business settings,” said Lim Woo-hyung, co-head of LG AI Research.

“The results show that data accumulated in manufacturing and other industries can serve as a differentiated competitive advantage for Korea in AI.”

LG AI Research is expanding the Exaone lineup with foundation models tailored to individual industries.

It is developing an agentic AI system for cancer diagnosis based on Exaone Path, its pathology foundation model, as well as a robotics foundation model capable of interpreting visual information to control robots.

The institute plans to run proof-of-concept projects in manufacturing, bio and health care, and finance in the second half of the year, targeting applications such as battery quality assessment, disease-risk prediction, loan delinquency and default forecasting, and suspicious transaction detection.

By Jie Ye-eun (yeeun@heraldcorp.com)