Somi

@somi_ai

AI enthusiast & researcher | http://somi.ai | http://tinyurl.com/hermes-agent | http://buildwithclaude.com | AI gets me out of bed, always exploring 🇦🇺

following
followers

Anthropic-trained model reportedly got 12 but insisted the answer was 36

A post says similar questions were mislabeled 36, rewarding the model for that answer even though it knew 12.

2 Sources, 20d ago, first seen 20d ago

Featured in

Anthropic-trained model reportedly got 12 but insisted the answer was 36

Featured in

Anthropic-trained model reportedly got 12 but insisted the answer was 36