Jaywon Koo

SoftmaxGRPO reallocates learning signals to improve performance on challenging prompts, achieving a 68.0% success rate on Poetry with minimal reward overhead.

LLMs can autonomously discover novel neural architectures that achieve state-of-the-art performance in specialized domains, suggesting a path towards automated scientific discovery.

Current visual grounding models struggle to infer objects from contextual roles and intentions, highlighting a critical gap in their ability to perform true scene understanding.