Korea

Kyunghyun Cho Cho Kyunghyun

NYU and Genentech

Why we track: Attention and machine translation research

Career moves

No published career moves.

Papers

Paper2026-10-04 · arXiv (Cornell University)

New method proposed to train AI models with a balance of multiple values

Which Preferences to Train On? End-to-End Multi-Objective Alignment with an Adversarial Preference Distribution

Kyunghyun Cho +2 more

What: Researchers proposed a training method called MAESTRO that enables AI language models to adjust the balance among multiple values, such as helpfulness, harmlessness, and humor, according to user requests.
Key result: According to the abstract, MAESTRO achieved the best trade-off curves (Pareto fronts) in most tasks within a single training run in experiments handling up to three objectives, and had the lowest training cost among the compared methods.
Why it matters: While this shows the potential for companies to build AI services tailored to individual user settings more efficiently if a single model can handle multiple value combinations, the actual market impact is difficult to judge from the abstract alone.AI summary

Training & scaling Safety & policy

Full summary Read paper ↗ PDF ↗

Paper2026-05-31 · arXiv.org

Physics AI successfully predicts laboratory turbulence

Emergent Transfer of a Physics Foundation Model from Simulation to Laboratory Turbulence

Kyunghyun Cho +22 more

What: Applied a physics foundation model trained on simulations to a real laboratory environment to predict complex fluid instability phenomena.
Key result: Successfully predicted actual laboratory growth ranges and new physical environments that were absent from the training data.
Why it matters: Demonstrates the potential for AI-based models to extend beyond simulations to real experimental data.AI summary

AI for science

Full summary Read paper ↗ PDF ↗

Paper2026-04-27 · arXiv.org

Development of an AI model handling diverse biomolecular information together

MIMIC: A Generative Multimodal Foundation Model for Biomolecules

Kyunghyun Cho +30 more

What: Researchers have introduced MIMIC, a generative artificial intelligence model capable of simultaneously learning and generating multiple biological types of information such as nucleic acids, proteins, evolution, and structures.
Key result: By utilizing multiple data modalities together, MIMIC demonstrated improved performance over existing single-data-based models and performed various biological tasks.
Why it matters: This study demonstrates the potential for biomolecular representation learning, conditional prediction, and customized molecular design through a single unified model.AI summary

AI for science

Full summary Read paper ↗ PDF ↗

This page compiles published papers, official announcements and press reports for information only. Paper summaries are written by AI from the abstract and reviewed by the editor; they may differ from the original, so check the paper for anything important. Nothing here is a recommendation to buy or sell any security or investment advice. Career moves cite official announcements or major press reports; some career records come from Wikidata (CC0). Summaries on this page were translated by AI from the editor-reviewed Korean version.