Nayoung Kim

Current PXRD-to-structure models achieve a 44.2% Top-1 match rate on experimental data, but still struggle significantly with generalization to unseen structural prototypes and recovery from two-phase mixtures.

Human operators are not static resources; adaptive task allocation that accounts for real-time workload and fatigue significantly boosts multi-robot supervision performance.

Packora achieves unmatched performance in molecular crystal structure prediction, outperforming traditional methods in both generation and ranking tasks.

Adaptive triggering can recover lost accuracy in LLM reasoning while cutting down on unnecessary interventions, challenging traditional fixed-interval approaches.

Heterogeneous LLM collaboration boosts robotic task planning success rates by ensuring plans are both feasible and efficient.

Forget expensive MoCap suits: this $150 force-feedback mouse lets you teleoperate dexterous robot hands with high task completion rates and reduced user workload.