Attendees visit the KISTI exhibition booth at KDD 2026 to watch a service demonstration. [KISTI]
Attendees visit the KISTI exhibition booth at KDD 2026 to watch a service demonstration. [KISTI]

An AI technology capable of analyzing vast bodies of scientific literature to generate new research hypotheses — and significantly reducing administrative burdens on researchers — has been unveiled for the first time.

The Korea Institute of Science and Technology Information (KISTI) announced Friday that it showcased a range of core technologies at KDD 2026, the world's leading academic conference on data mining and AI. The technologies on display included a science-focused large-scale AI model, an autonomous research platform, an AI agent and a large language model (LLM) development platform.

The centerpiece of the exhibition was "AI Scientist," a system designed to support the scientific discovery process itself. It uses AI to search and analyze large volumes of scientific literature, derive new research hypotheses and suggest directions for further inquiry. The goal is to reduce the time and effort researchers spend manually searching and reviewing papers while helping them generate new research ideas.

Also on display was "Web Agent," an AI agent built to automate research administration tasks. The system reads information from files such as a researcher's curriculum vitae and research proposals, then automatically fills in the required fields across various web-based forms. By handling repetitive administrative work, the agent aims to free researchers to focus on their actual research.

"KONI-Forge," an LLM development platform built on high-performance computing (HPC) infrastructure, integrates the entire development pipeline — from fine-tuning and model evaluation to optimization and service deployment. The platform is designed to help researchers and companies develop and apply AI models tailored to science and technology fields more efficiently.

Alongside this, KISTI presented a science-specialized multimodal language model (MLLM) platform, high-performance fine-tuning technology and a large-scale scientific dataset preprocessing pipeline. The institute highlighted its progress in building an open-science AI ecosystem for joint use by academia and industry, and expanded technical exchanges with global researchers and corporate representatives.

"The KDD 2026 exhibition provided an opportunity to introduce KISTI's 'AI for Science' capabilities, autonomous agent technologies and LLMOps platform achievements to the global data and AI research community," said Lee Gyeong-ha, head of KISTI's Large-Scale AI Research Center. "We will continue to advance AI Scientist, intelligent agents and the KONI-Forge platform to lead a national science and technology AI ecosystem and drive global research innovation."


nbgkoo@heraldcorp.com