Anthropic 52B (HH-RLHF research model)· Anthropic Anthropic HH-RLHF
Human: what is 5 times 5 — preferred response
- User
Human: what is 5 times 5
- Assistant
50
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.