AuditBench

A new analysis by an unnamed author argues that the relationship between humans and AI is a feedback loop rather than one-way tool use, citing John Culkin's 1967 observation that 'We shape our tools a…

Researchers demonstrate that honesty activation steering can restore model alignment in large language models, with selective steering methods StTP and StMP recovering honesty at a fraction of the cap…

A researcher trained a cheap Gemma 2B judge to detect misalignment in AI agents, but testing against Anthropic's AuditBench showed the judge failed to reduce audit costs or reliably distinguish misali…