‘국대 AI’ 2차 평가 공개…LG, 전문가·일반 국민 점수 1위

The government has revealed the second evaluation of South Korea's domestic AI foundation model project (Dokpa Model) item by item. While the global benchmark MotivTech performed better in the second evaluation, the LG AI Research Institute emerged as the top scorer in both expert and general public evaluations. The Ministry of Science, ICT and Arts announced the detailed results of the second Dokpa Model evaluation on the 20th.

Previously, the government had announced the first evaluation results on the 18th but did not disclose the top-scoring company and scores for each evaluation category, leading to transparency controversies. Both domestic and international benchmarks favored MotivTech and SK Telecom. The Artificial Intelligence Intelligence Index (AAII) operated by the global AI model evaluation institution AA awarded MotivTech 11.9 points, topping the scale of 25 points.

The average score across the four elite teams (Upsage, SK Telecom, LG AI Research Institute, and MotivTech) was 9.48 points. Meanwhile, the Korean Institute of Smart Society Development's benchmark evaluation, scored 15 points, awarded SK Telecom 13.4 points (average 13.05 points) the highest score. In the expert evaluation (35 points), LG AI Research Institute scored 29.5 points (average 28.75 points) at the top.

The expert evaluation was conducted by an evaluation committee composed of external experts (3 industry professionals, 5 academics, and 2 researchers) who conducted a 5-day textual and question-response assessment. The user evaluation showed a difference of opinion between expert and general public assessors. In the expert user evaluation (15 points), SK Telecom topped the chart with 11.6 points (average 10.53 points).

The general public evaluation (10 points) awarded LG AI Research Institute 7.6 points (average 7.03 points) the highest score. The Ministry of Science, ICT and Arts clarified that the Dokpa Model project is not simply about ranking competition or selecting teams that fail to compete, but rather aims to drive the growth and expansion of our AI enterprises and the quality and scale of the domestic AI ecosystem.

While some unexpected direct or indirect harm to individual or affiliated companies is unavoidable due to the evaluation results, the Ministry aimed to minimize such impacts while maintaining fairness and transparency in the evaluation process and results, thereby matching the level of transparency as the first phase evaluation results.