Taekyung Kim

Instantaneous geometric scans leave robots blind to cascading physical interactions, but grounding VLM physical intuition with conformally certified backup control vaults safety rates from 43.3% to 99.3%.

Standard pixel augmentations frequently corrupt delicate vision-language alignment, but injecting diffusion-style isotropic noise directly into embedding spaces breaks through the longstanding performance ceiling of stacked CutMix, Mixup, and RandAug recipes.

Chain-of-thought is far more than surface imitation: LLMs carve out discrete, context-dependent geometric subspaces for functional operations like deduction and decomposition, with middle layers encoding the abstract reasoning step rather than the literal tokens.