The Ministry of Science and ICT has released the detailed results of the second evaluation phase of the Independent AI Foundation Model (Dukpamo) project. SK Telecom (SKT) and LG AI Research Institute each ranked first in two evaluation categories, while Motif Technologies, which joined through an additional application, scored the highest in one category but was eliminated in the second evaluation.
On August 20, the ministry disclosed the detailed scores and average scores for the five evaluation categories of the second phase.
In the evaluation results, Motif achieved a score of 11.9 in the Artificial Analysis (AA) AI index evaluation, ranking first. The average score among the four teams was 9.48. In the NIA benchmark, which assessed seven categories including mathematics, knowledge, long text comprehension, safety, reliability, Korean language, and instruction execution, SKT received the highest score of 13.4.
In the expert evaluation, which allocated a total of 35 points, LG AI Research Institute secured first place with a score of 29.5. The average score among the four teams was 28.75.
In the AI specialist user evaluation, which had a maximum score of 15 points, SKT achieved the highest score of 11.6, with an average score of 10.53.
In the newly introduced public evaluation for the second phase, LG AI Research Institute also ranked first.
The ministry stated, "The Dukpamo project is not merely a ranking competition or a means to eliminate teams, but aims to foster growth among our AI companies beyond competition and to drive qualitative growth and expansion of the domestic AI ecosystem." It added, "We will expand policy support to ensure that the elite teams of Dukpamo and various companies do not stop at the results of this project but continue to develop their AI models and technologies for practical use and dissemination in the field."
Meanwhile, the ministry announced on August 18 that it is reviewing plans to reorganize the competitive support system for the Dukpamo project. The third phase evaluation will proceed as originally planned, but the continuation of the scheduled fourth phase evaluation will be subject to further review. Specific directions for developing frontier-level AI models linked to the Dukpamo project will be announced soon.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.