Menu
Inshorts
For the best experience use inshorts app on your smartphone
inshortsinshorts
AI researcher Lun Wang exits DeepMind, warns of benchmark gaps
short by / on Tuesday, 19 May, 2026
AI researcher Lun Wang has departed Google DeepMind, warning that static benchmarks cannot accurately measure advanced large language models. He noted that models often memorise tests or exploit patterns, creating dangerous safety and capability gaps. To combat this, Wang proposes "self-evolving evals" adaptive testing systems that dynamically update to ensure real-world reliability and safety.
read more at BreezyScroll