A robot passes a thread through a narrow needle eye with ease.

It moves smoothly without colliding with obstacles, even when the clearance is only 2.5mm.

The robot performs delicate and precise movements that would require intense concentration from a human, all without a hitch.

While existing methods required vast amounts of data and training for such precise movements, researchers at KAIST have developed a technology that enables learning with significantly less data.

The research team trained a model that learns changes over time and another that generates various actions. By doing so, they developed a robot that demonstrates superior performance even when trained on only one-quarter of the data used in previous studies.

The team explained that while previous research treated every moment of human movement with equal importance, this study focused less on large movements and placed greater weight on learning precise actions, such as manipulating a thread.

The robot's performance also showed a significant difference.

Experimental results showed a task success rate up to 81% higher than that of conventional robots.

[Interview] Park Dae-young / Professor, School of Computing at KAIST: "Because we now take dynamics into account, the robot understands and learns the difference between fast and slow. Since it has shown high performance in areas requiring very precise control, we expect it to be applied to sectors in our country that have not yet been automated."

As the technology reduces the amount of data required for training, it can also help address issues related to GPU and memory shortages.

The research team expects this technology to be applied in various industrial fields that require high accuracy, such as precision parts assembly, manufacturing, and medical surgery.

"Navigating Needle Eyes Freely".. Less Data, Higher Success Rate (June 24, 2026, D-Report)

Reported by Seo Dong-kyun | Written by Lee Seyoung | Video by Choi Hye-young | Produced by SBS Digital News