Researcher in AI safety with a PhD in AI from UC Berkeley.
Based on 7 scored articles
Referenced in coverage
“Co-author of the MATH dataset paper measuring mathematical problem solving with language models.”
“Co-author of MAIM proposal for deterrence strategy in superintelligence development.”
“Dan Hendrycks authored the Safetywashing paper, which Alan Chan referenced in discussing model-benchmark level safety claims.”
“Creator of MMLU benchmark who promised to develop MMLU Pro to address benchmark saturation issues.”
“Dan Hendrycks discussed skepticism about mechanistic interpretability's past emphasis in AI safety research.”
“Led a team including Yoshua Bengio that published an AGI framework and metrics dividing general intelligence into 10 cognitive domains.”
“Lead author of MMLU benchmark; co-authored Humanity's Last Exam (HLE), a more difficult benchmark.”