Selectstar's 3 AI Papers Accepted at EMNLP 2025

## SelectStar Proves Global Leadership in AI Reliability Technology…Numerous Papers Accepted at EMNLP 2025 SelectStar, a company specializing in AI data and reliability verification, has once again proven its global technological prowess by having an impressive three papers accepted at 'EMNLP 2025', one of the world's top three natural language processing (NLP) conferences. Notably, one of these papers achieved the remarkable feat of being accepted as a main paper, placing it in the top 22%. This is noteworthy as it solidifies SelectStar's unique technological leadership in the fields of AI data bias minimization and reliability verification on the international stage. EMNLP, the most prestigious conference under the Association for Computational Linguistics (ACL), boasts fierce competition with over 8,000 papers submitted and only 39.51% accepted. On such a stage, SelectStar successfully had an additional three papers (for a total of five) accepted this year, following two last year, clearly demonstrating its AI research capabilities. The papers accepted this time focused on solving fundamental problems in AI. The main paper, 'CoBA', presented an automated text augmentation technology based on Large Language Models (LLMs) that dramatically reduces bias in training data, simultaneously enhancing the fairness and performance of AI models. Additionally, the 'GRADE' paper proposed a multi-hop question-answering evaluation framework capable of precisely diagnosing the inference and retrieval capabilities of RAG systems, while 'CAC-CoT' introduced an innovative synthetic data generation technique that resolves the inefficiencies of existing inference methods, maximizing data efficiency. Through these contributions, SelectStar has helped solve core challenges such as AI reliability, data bias minimization, and advanced inference performance. These core technologies will not only be utilized in generating high-quality data for national AI projects but will also be actively integrated into 'Datumo Eval', SelectStar's self-developed AI reliability verification SaaS. This is expected to contribute to the establishment of reliable AI systems across various industries. Kim Se-yeop, CEO of SelectStar, stated, “The acceptance at EMNLP signifies official recognition from the global academic community for SelectStar’s unique AI reliability verification technology,” adding, “Moving forward, we will further solidify our technological leadership in AI reliability verification based on these research achievements, ultimately taking the lead in creating a trustworthy AI ecosystem.” Meanwhile, SelectStar is already leading the domestic AI reliability verification market by successfully supplying its self-developed AI reliability verification solution, 'Datumo Eval', to major financial institutions and various other industries within Korea.