Motif Technology, which was eliminated from the second-stage evaluation of the "Independent AI Foundation Model Project," has filed an objection to the review results. The company maintains that a specific explanation from the government is necessary, as it received the lowest overall score despite achieving the highest score in global benchmark evaluations.
On the 27th, Motif Technology issued a statement confirming that it has formally submitted an objection regarding the selection results of the second stage of the Independent AI Foundation Model Project. The project aims to build general-purpose AI models specialized for the Korean language and domestic industrial demand, with the final two models selected after three rounds of evaluation.
Although Motif joined the competition starting from the second round through an additional recruitment process following the first-stage evaluation, it was ranked last among the four participating teams (Upstage, SK Telecom, LG AI Research, and Motif Technology) in the evaluation held on the 18th and was eliminated.
Initially, Motif stated that it would respect and accept the evaluation results. However, it appears the company proceeded with an objection after negative assessments regarding the performance of its own model, Motif 3, spread through media reports.
Motif stated, "While we respect and accept the evaluation results themselves, subsequent media reports indicated that 'Motif 3' was described as having high scores only on AAII benchmarks but facing issues in expert and user evaluations." The company further noted, "High-level officials from the Ministry of Science and ICT also reportedly commented that Motif 3's actual reasoning performance is insufficient compared to its benchmark scores, and its usability and applicability are lacking."
The company continued, "This goes beyond simply determining which companies were selected or eliminated; it raises fundamental questions about the essence of the Independent AI Foundation Model Project and the direction it should take in the future." It added, "We also feel there may be some misunderstandings regarding the technical direction we have pursued and its outcomes," and formally requested a detailed release of evaluation scores and a re-evaluation.
Furthermore, Motif stated, "The value of foundation models lies not in chatbot usability but in foundational performance that can be extended across multiple domains," emphasizing that "when a model's intrinsic performance is high, its potential for expansion develops rapidly." This serves as a rebuttal to the criticisms regarding 'usability' and 'applicability,' which were cited as reasons for Motif 3's elimination.
The company also noted, "While we recognize that our focus on intrinsic performance may differ from factors some experts consider important, it remains to be seen which direction is more appropriate, as this will be determined through future technological advancements and market validation."
Additionally, Motif highlighted that its model, Motif 3, achieved the highest score of 47 points among the four companies in the Independent AI Foundation Model Project on AAII, a global comprehensive performance indicator (Upstage: 37 points, SKT: 35 points, LG AI Research: 31 points). The company stressed, "Despite the significant difference of 47 points versus 31 points between models, the reasons and background for such contradictory results must be explained more concretely so that companies and the industry can understand." AAII is not a single benchmark but a comprehensive evaluation aggregating nine representative benchmarks widely used in the global AI industry; therefore, it should be sufficiently considered as an important reference standard when judging performance differences between models.
Motif concluded, "Establishing a rational direction for the Independent AI Foundation Model Project is extremely important at the national level," adding, "The review process and results of this round must be fully disclosed and verified, so that in the third round, participating companies can focus their capabilities under clearer directions."
in TECH
