Metilience unveils a hybrid AI reasoning engine for high-stakes exams, leveraging structured cognitive error analysis via its newly launched Credic platform.
SEOUL, SEOUL, SOUTH KOREA, February 27, 2026 /EINPresswire.com/ — Metilience, an AI education technology startup, announced the launch of a hybrid reasoning engine designed to analyze recurring cognitive errors in high-stakes academic and professional exams, including LSAT-level logical reasoning environments.
The platform combines retrieval-augmented reasoning (RAG) with selective parameter-efficient fine-tuning experiments conducted on proprietary cognitive error metadata compiled since 2021. According to the company, its dataset consists of anonymized reasoning breakdown structures and structured trap classifications rather than copyrighted exam content.
Rather than focusing solely on answer memorization, Metilience aims to model how reasoning errors repeat under time pressure. The company states that identifying consistent cognitive blind spots allows for more targeted intervention compared to traditional repetition-based preparation models.
The architecture integrates a base large language model accessed via API, a structured reasoning error taxonomy, a vector database of anonymized cognitive metadata, adaptive recommendation logic, and domain-specific fine-tuning modules. The platform is independently developed as part of the company’s ongoing research into structured cognitive error analysis.
Elements of the reasoning framework were informed by formal logic studies completed through Stanford Online. Metilience notes that its approach draws on structured logic principles to categorize misinterpretation patterns, conditional reasoning failures, and trap-type misclassification tendencies frequently observed in standardized testing environments.
The first commercial deployment of the technology is Credic, a subscription-based structured test preparation platform currently operating in South Korea. Credic applies the company’s cognitive error mapping framework to English-language standardized testing environments and delivers personalized daily reasoning exercises based on identified trap patterns.
According to Metilience, Credic’s recommendation logic adapts dynamically to user performance data, adjusting problem selection based on recurring reasoning errors rather than simple difficulty progression. The company describes this as an attempt to move from content-centric preparation to pattern-centric reasoning training.
Metilience clarified that it is not affiliated with the Law School Admission Council (LSAC) and does not use official LSAT materials. The company states that its reasoning engine is designed for graduate-level and professional entrance-level logical reasoning preparation contexts more broadly.
The company was founded by Gwangseok “Keith” Kim, a former medical student currently studying at Korea University. His academic background in structured reasoning contributed to the early development of the platform’s error-classification framework.
According to Metilience, its long-term objective is to build a cross-exam reasoning infrastructure capable of supporting multiple high-stakes examination categories, including graduate admissions and professional qualification environments.
Metilience has not disclosed funding details and continues refining its AI education technology platform through structured experimentation, model iteration, and expanding its proprietary cognitive metadata framework.
Metilience – https://metilience.com
Credic – https://metilience.com/credic
Media Contact:
press@metilience.com
Metilience Press Office
Metilience
official@metilience.com
Legal Disclaimer:
EIN Presswire provides this news content “as is” without warranty of any kind. We do not accept any responsibility or liability
for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this
article. If you have any complaints or copyright issues related to this article, kindly contact the author above.
