Ranked No. 1 overall on TabArena, ahead of Google Tops Salesforce in predictive evaluation as well Strong in defect detection and financial risk prediction PoC launches in manufacturing, biotech, and finance in H2
LG’s artificial intelligence (AI) model “EXAONE” has outperformed U.S. and Chinese big tech companies in a global assessment of industrial data prediction, achieving world-leading performance. Observers say this demonstrates that LG’s technological edge is recognized not only in the competition among general-purpose language models, but also in industrially specialized AI that can be applied directly to corporate operations.
LG AI Research Institute announced on the 7th that “EXAONE Tabular,” which analyzes tabular data, and “EXAONE Forecast,” which predicts future data based on time-series trends, have each recorded state-of-the-art (SOTA) performance on their respective global leaderboards. These global leaderboards are international evaluation platforms that benchmark the predictive capabilities of AI models using real-world industrial data.
Leaderboard screen showing LG AI Research Institute’s “EXAONE Tabular” ranked first overall in “TabArena,” a global assessment of tabular data analysis. EXAONE Tabular achieved a score of 1,760 points, surpassing Google’s latest model (1,749 points) and proving its state-of-the-art (SOTA) performance.
EXAONE Tabular, which handles spreadsheet-like data organized in rows and columns, took first place in the “categorical data prediction” category of the related global evaluation (TabArena). The model scored 1,760 points, beating Google’s latest model, unveiled last month, which scored 1,749 points. It can identify defective products in manufacturing processes or rapidly predict risk factors in the financial sector using only small amounts of data, offering high practical utility on the ground. Even when portions of the collected data are missing, it can accurately fill in the gaps by analyzing surrounding information.
EXAONE Forecast, which predicts data that change over time, also claimed the top spot in a prediction benchmark led by global company Salesforce. In the category that evaluates the accuracy of predictions for data from domains the AI encounters for the first time, without additional training, it ranked first ahead of Google and Alibaba. It can track complex, constantly changing indicators—such as product demand, raw material prices, and stock prices—with a single general-purpose AI model.
Lim Woo-hyung, Co-Head of LG AI Research Institute, stated, “This is a case that proves the data accumulated at various industrial sites, including manufacturing, can become a differentiated competitive advantage for Korean AI.” Starting in the second half of this year (July–December), LG plans to conduct proof-of-concept (PoC) projects for these technologies in manufacturing, bio and healthcare, and financial sectors, and to link them to actual business.
Kim Jae-hyung
AI-translated with ChatGPT. Provided as is; original Korean text prevails.
ⓒ dongA.com. All rights reserved. Reproduction, redistribution, or use for AI training prohibited.
