Report
GPT-6.1 Sol, Opus 5.5 and Gemini 4 Argon reportedly identify most of the frontier-lab researchers tested by name
A user says Argon is much less reliable, often naming researchers' former employers as their current ones.
TLDR
A user reports that GPT-6.1 Sol, Opus 5.5 and Gemini 4 Argon identified most of the tested frontier-lab researchers by name. The user says the models usually knew the researchers' areas of work and previous work, but Argon often named a former employer as the current one.
Combined views
—
3 Sources, first seen ago
