3 ms·
Anthropic/Claude does not use any RLHF.
by MacsHeadroom 2y ago
Anthropic/Claude does not use any RLHF.
- cjbillington 2y agoWhat do they do instead? Given we're not talking to a base model.
- tqi 2y agoSupposedly they use "RLAIF", but honestly given that the first step is to "generate responses... using a helpful-only AI assistant" it kinda sounds like RLHF with more steps. https://www.anthropic.com/research/constitutional-ai-harmlessness-from-ai-feedback https://www.anthropic.com/research/constitutional-ai-harmles...
- teqsun 2y agoIs that a claim they've made or has that been externally proven?