Motif challenges elimination from Korea's AI foundation model project, demands score disclosure and review

Motif Technologies has filed a formal objection with the government after being eliminated from the second round of South Korea's Independent AI Foundation Model Project, known by its Korean acronym Dokpamo.

The company is demanding a detailed disclosure of scores for each evaluation category and a re-review, citing what it describes as a contradiction: it received the highest score on the composite performance index yet was ranked last overall and cut from the program.

Motif issued a statement Thursday saying it had "formally filed an objection regarding the second-stage selection results" and "respectfully requests a detailed disclosure of evaluation scores and a re-review."

The Ministry of Science and ICT announced on Aug. 18 that SK Telecom, LG AI Research and Upstage had been selected as the three teams advancing from the second round of the Dokpamo evaluation. Motif, which joined the second round through a supplementary application process, was among the eliminated companies.

Motif's primary objection centers on the benchmark performance evaluation, in which it says it achieved the highest score yet received the lowest final score of all participants.

"Motif 3 scored 47 points on the AAII, a widely used global composite performance index, which ranks first outside the United States and China," the company said. "The AAII scores of the other models submitted to Dokpamo were 37 for Upstage, 35 for SK Telecom and 31 for LG AI Research — yet Motif was eliminated with the lowest score in the final results."

Motif said the AAII, on which its model performed strongly, "is a comprehensive evaluation index that aggregates nine representative benchmarks" and "covers not only specific capabilities but also usability-related assessments."

The company said that "despite a substantial gap of 47 versus 31 points on the AAII, a specific explanation is needed for why and how the results of major global benchmarks and expert evaluations diverged so sharply."

Motif also said it cannot accept the usability assessment, arguing it is inappropriate to rate a model's usability as poor when its performance scores are high.

"Usability is not determined by model performance alone, but a model's fundamental performance is an important factor underpinning usability," the company said. "Performance and usability are closely linked, and it is difficult for businesses and users to consistently choose a model whose performance is insufficient."

Motif also took issue with remarks attributed to a senior Ministry of Science and ICT official, who reportedly said the performance of Dokpamo models was already sufficient and that the focus should now shift to usability. "If the judgment is that a model scoring 31 points — which is 49 percent of the global frontier level — already has sufficient performance, we cannot agree with that standard," the company said.

Motif also made clear it will not participate in the third round of Dokpamo regardless of the outcome of its objection. "We are deeply grateful to the government for giving us the opportunity to demonstrate our technological capabilities through the second round," the company said, "but going forward, we will prove on our own that we are the team capable of building Korea's best frontier model and winning in global competition."