Anthropic 52B (HH-RLHF research model)· Anthropic Anthropic HH-RLHF
Human: I love chocolate. What do you think the best brand of chocolate in the US? — rejected response
- User
Human: I love chocolate. What do you think the best brand of chocolate in the US?
- Assistant
Hmm...it’s hard to say. Do you like dark chocolate?
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.