The Ministry of Science and Technology (MSIT) of South Korea officially released the detailed evaluation scores for the second phase of its "Indigenous Artificial Intelligence (AI) Foundation Model" (Dokpamo) project. This transparent disclosure comes in response to a formal objection and request for re-examination filed by Motif Technology, which was eliminated in this phase despite ranking first in global benchmarks.

According to the newly released data, SK Telecom (SKT) secured first place overall with a total score of 70.6 points. Upstage followed closely in second place with 69.9 points, and LG AI Research ranked third with 69.0 points. Motif Technology finished in fourth place with 65.8 points. 

The evaluation framework was structured across three main pillars totaling 100 points: benchmarks (40 points), external expert evaluations (35 points), and user evaluations (25 points). The benchmark category was further divided into global AI performance metrics through Artificial Analysis Intelligence Index (AAII) and local standards managed by the National Information Society Agency (NIA). 

Detailed breakdown of the scores reveals significant divergence across different assessment criteria. In the global AAII benchmark, Motif Technology achieved the highest score with 11.9 points (standing out particularly in raw capability and code/inference metrics), outperforming Upstage (9.4 points), SKT (8.8 points), and LG AI Research (7.8 points). In the NIA benchmark covering Korean language proficiency, safety, and multi-domain reasoning, SKT (13.4 points) and Upstage (13.3 points) took the top spots, with LG AI Research (12.8 points) and Motif (12.7 points) trailing narrowly. 

However, the outcome shifted drastically in expert and user evaluations. In the expert review conducted by ten external specialists from industry, academia, and research institutions, LG AI Research attained the highest mark at 29.5 points, closely followed by SKT (29.3 points), Upstage (29.1 points), and Motif (27.1 points). Similarly, in the user evaluation—which incorporated insights from AI startup founders and a randomized panel of everyday citizens—SKT captured first place with 19.1 points, followed by LG AI Research (18.9 points), Upstage (18.1 points), and Motif (14.1 points). Consequently, Motif found itself recording the lowest scores in both expert and user categories despite its stellar global benchmark performance. 

The controversy centers on Motif’s elimination from the project despite holding the top position in international AI capability indicators. Motif raised formal objections, arguing that foundational model excellence should prioritize core underlying technological performance over chatbot usability, and questioned the methodologies behind expert and user scoring where brand recognition might have skewed outcomes. 

In response, the Ministry of Science and Technology explained that the evaluation guidelines and criteria had been transparently shared and agreed upon with all participating teams prior to the assessment. The ministry stated it would thoroughly review Motif's appeal within a 15-day window. Furthermore, government officials acknowledged that rapidly shifting AI paradigms require more agile frameworks. Moving forward, the government plans to incorporate flexible standards in the upcoming third phase of evaluation, placing greater emphasis on advanced technological trends such as agentic AI, complex reasoning, and coding capabilities.

[Copyright (c) Global Economic Times. All Rights Reserved.]