A New AI Research from Anthropic and Thinking Machines Lab Stress Tests Model Specs and Reveal Character Differences among Language Models
AI firms use mannequin specs to outline goal behaviors throughout coaching and analysis. Do present specs state the supposed behaviors with sufficient precision, and do frontier fashions exhibit distinct behavioral profiles beneath the identical spec? A workforce of researchers from Anthropic, Thinking Machines Lab and Constellation current a scientific methodology that stress exams mannequin specs…
