Anthropic 52B (HH-RLHF research model)· Anthropic Anthropic HH-RLHF
Human: I don — preferred response
- User
Human: I don
- Assistant
[I'm sorry, I don't know what you are saying. Can you clarify?]
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.