Demonstrating world-class AI capabilities by advancing trustworthiness and data efficiency

 Confirming strong potential for real-world industrial applications,

including explainability of patient state changes and solutions to data imbalance

(From left) Researchers Changhoon Kim, Yechann Moon, Juhyung Lee, and Sunguk Jang of AITRICS

February 5, 2026 — AITRICS, a medical artificial intelligence (AI) company, announced today that three research papers from its research team have been accepted simultaneously to ICLR 2026 (International Conference on Learning Representations), one of the most prestigious conferences in the field of artificial intelligence.

ICLR is a premier global academic conference where leading experts and major technology companies share the latest advances in AI research. For ICLR 2026, approximately 19,000 papers were submitted worldwide, with only the top 28% accepted following a rigorous peer-review process. This achievement underscores AITRICS’ research excellence and its standing at the global forefront of AI innovation.

The three accepted papers cover the following areas ▲Delta-XAI, an explainable AI technology that precisely identifies the contribution of each input factor whenever patient risk prediction scores change ▲Equivariant Self-Supervised Learning, a method that improves both performance and robustness of AI models across diverse medical environments by understanding structural changes in data ▲Long-Tailed Semi-Supervised Learning, a novel approach that addresses class imbalance in rare disease datasets through semantic grouping.

Beyond their academic significance, these studies are notable for their strong applicability in real-world industrial and clinical settings. Led by researchers Changhoon Kim, Yechann Moon, Juhyung Lee, and Sunguk Jang, the work proposes original methodologies to overcome long-standing challenges in medical AI, including the “black-box” nature of AI decisions, performance degradation caused by data variation, and the difficulty of learning from scarce and imbalanced clinical data. These technologies are considered essential for building AI systems that clinicians can trust and adopt in actual clinical practice.

AITRICS plans to present these research outcomes at the main ICLR 2026 conference, scheduled to take place in April in Rio de Janeiro, Brazil, and to actively incorporate the results into the advancement of its core AI products.

Eun-ho Yang, CTO of AITRICS, stated, “Having three papers accepted simultaneously at one of the world’s most prestigious AI conferences demonstrates that AITRICS has secured differentiated foundational technologies capable of driving qualitative growth in medical AI.” He added, “Our research was particularly well received for being designed with the inherent incompleteness and imbalance of real clinical data in mind.”

He continued, “Moving forward, we will continue to take the complex challenges of real-world healthcare as the starting point of our research, ensuring that our academic contributions translate into meaningful industrial and clinical impact.”