LG’s (003550) artificial intelligence platform “EXAONE” has surpassed AI from the United States and China in the field of industrial-specific AI foundation models, demonstrating world-leading technological capabilities. Its strategy based on high-quality industrial data is delivering differentiated results in the AI supremacy race dominated by Big Tech firms such as Google and Alibaba, according to industry assessments.
LG AI Research announced on the 7th that “EXAONE Tabular,” which analyzes and predicts tabular data, and “EXAONE Forecast,” which predicts future data over time, each recorded state-of-the-art (SOTA) performance on global AI evaluation platforms.
The axis of AI competition is shifting from a focus on large language models (LLMs) to “industrial-specific AI” that solves real industrial problems. EXAONE’s latest achievement carries significant meaning in that it reaffirms outstanding technological capability by leveraging data accumulated in the field across manufacturing, finance, and healthcare as a competitive advantage.
Lim Woo-hyung, co-head of LG AI Research, said, “Global Big Tech’s interest is rapidly shifting from general-purpose language models to industrial-specific AI foundation models that solve problems in actual industrial fields.” He added, “This case proves that data accumulated across various industrial fields, including manufacturing, can become Korea’s differentiated AI competitiveness.”
LG EXAONE Tabular
World’s Top Performance in Structured Data AI
Predicts Quality, Customers, and Bankruptcy Risk
LG AI Research said on the 7th that “EXAONE Tabular,” which analyzes and predicts tabular data, and “EXAONE Forecast,” which predicts future data over time, each achieved state-of-the-art (SOTA) performance on a global AI evaluation platform. LG AI Research
LG AI Research said EXAONE Tabular ranked first overall in the categorical data prediction category on “TabArena,” a leading global structured data AI evaluation platform. This proved that EXAONE Tabular has world-leading competitiveness in processing tabular data, the most widely used data type in industrial fields.
Serving also as a financial analyst, “EXAONE BI” — a solution based on LG’s large language model (LLM) EXAONE — is specialized in analyzing about 8,000 stocks listed on the Korean and U.S. markets each day. LG AI Research
Tabular data is the most broadly used data type in industrial settings, including manufacturing process data, customer information, and financial records. While existing AI models focused mainly on processing text and images, EXAONE Tabular was designed to directly understand data structures composed of rows and columns and the relationships between variables.
In particular, EXAONE Tabular showed top-level performance in both binary classification, which predicts two outcomes such as normal or defective, and multi-category classification, which predicts three or more outcomes.
Its evaluation score (ELO) was 1,760 points, surpassing the 1,749 points of “TabFM,” the structured data AI model recently unveiled by Google. In the structured data AI market where it competes with global Big Tech, LG took the lead in technological capability.
The strength of EXAONE Tabular is that it can quickly adapt to new problems with only small amounts of data. In industrial settings, it is often difficult to secure sufficient data when new products or processes are introduced, but EXAONE Tabular can derive prediction results while reducing the burden of additional training.
Tabular can also predict the likelihood of defects in manufacturing based on inspection data such as battery cell voltage and internal resistance. In bio-healthcare, it can analyze disease risk and treatment response using patient clinical data, and in finance, it can be used to predict the likelihood of delinquency, bankruptcy, and abnormal transactions. Furthermore, even when some data is missing, it performs highly accurate predictions by analyzing the relationships among the remaining data.
LG AI Research explained that Tabular has high potential for use in various fields, including anomaly detection in manufacturing sites, quality assessment, customer churn prediction, and financial risk analysis. Accordingly, the institute plans to conduct proof of concept (PoC) for EXAONE Tabular in manufacturing, bio-healthcare, and finance in the second half of this year and connect it to actual commercialization.
LG “EXAONE Forecast”
No. 1 in Future Prediction AI, Beating Google and Alibaba
EXAONE Forecast, another core model from LG AI Research, recorded world-leading performance in the field of time-series AI that predicts future data.
LG AI Research on the 31st unveiled “K-EXAONE 2.0,” the second model in the “Sovereign AI Foundation Model Project” led by the Ministry of Science and ICT. LG
LG AI Research said EXAONE Forecast ranked first in the world in the zero-shot category on “GIFT-Eval,” a time-series forecasting evaluation platform developed by global customer relationship management (CRM) company Salesforce.
Zero-shot is a metric that evaluates AI’s ability to make predictions on data it encounters for the first time without additional training. EXAONE Forecast recorded the top performance, beating models from global companies such as Google and Alibaba. It also ranked second in the “Agentic AI” category, in which AI analyzes data and performs predictions on its own in an actual work environment, demonstrating both its applicability to industrial fields and its versatility.
EXAONE Forecast was trained on real data from various fields such as internet, sales, energy, economy, healthcare, and transportation, as well as 2 trillion virtual time-series data reflecting the real world. Through this, it was designed to predict various future scenarios by comprehensively assessing trends, seasonality, volatility, structural changes, and interactions between variables, going beyond simply analyzing past patterns.
LG AI Research is already using time-series prediction AI in the field at affiliate companies. Together with LG Electronics (066570) and LG Energy Solution, it developed and applied models to predict seasonal product demand and raw material prices such as lithium.
Starting this year, it has also launched a financial AI service that analyzes about 8,000 stocks listed on the Korean and U.S. markets in cooperation with Koscom and the London Stock Exchange Group (LSEG). With AI providing prediction scores and analysis opinions for each stock, it is expanding its scope of use into the financial forecasting domain.
EXAONE, Beyond General-Purpose AI to Medicine and Robots
Expanding into Industrial-Specific AI Foundation Models
LG AI Research is accelerating its efforts to develop EXAONE not as a simple language model but as an industry-specific AI platform.
An AI-generated image depicting points of connection in the AI era. Seoul Economic Daily DB
A representative example is the medical field. LG AI Research unveiled “EXAONE Path,” a pathology foundation model, in 2024. It is currently developing “Cancer Agentic AI” to help with cancer diagnosis and medical staff’s decision-making. It has also begun developing a “Robot Foundation Model (RFM)” that enables AI to understand visual information and perform tasks in actual physical spaces.
LG’s strategy is to expand EXAONE’s scope of application based on data secured in industrial fields such as manufacturing, finance, medicine, and robotics, and to strengthen Korean-style industrial AI competitiveness.
Lee Soon-young, head of the Data Intelligence Lab at LG AI Research, said, “LG’s differentiated competitiveness is the data accumulated across various industries and its deep understanding of industrial fields.” He added, “We will continue to expand our group of foundation models specialized for various industrial fields and create a new axis of competition in the global AI market.”