Agentic artificial intelligence startup Enhance said on Wednesday that it had three papers accepted at three international AI conferences.

It said all three were led by Enhance AI researcher Won-deok Seo (서원덕) as first author. Enhance plans to use the achievement to advance its core AgentOS service technology.

A paper co-researched with Peking University and Fudan University was accepted at the International Conference on Machine Learning (ICML 2026). It proposed the OG-MAR framework, which structures demographically similar respondent profiles in an ontology format to address cultural bias in large language models.

A paper on automated data science co-researched with Yale University and Peking University was accepted at the Association for Computational Linguistics (ACL 2026). It proposed the multi-agent-based SPIO framework, which explores and ensembles multiple planning paths to build a more flexible and robust analysis pipeline than before.

A visualization code synthesis paper co-developed with UC Berkeley, Princeton University and Peking University was accepted at the International Conference on Pattern Recognition (ICPR 2026). It included the VisPath framework, which explores visualization directions through multiple reasoning paths even for ambiguous natural-language requests and improves the final code by reflecting visual feedback.

Enhance CEO Seung-hyun Lee (이승현) said, "Having three papers simultaneously accepted at the world's most authoritative conferences across diverse AI fields including natural language processing, vision and machine learning proves Enhance's leading research capabilities." He added, "We will continue to expand global joint research with researchers at world-class universities and set the standard for next-generation AI agent technology."