TLDR AI · RESEARCH & INNOVATION
SABOTAGE EVALUATIONS FOR FRONTIER MODELS
Summary from the TLDR AI newsletter
Anthropic has performed a number of novel evaluations to determine weaknesses and how to measure misalignment in large and powerful models.
TLDR AI · RESEARCH & INNOVATION
Anthropic has performed a number of novel evaluations to determine weaknesses and how to measure misalignment in large and powerful models.