4 ms·
Anthropic scores fine-tuning in the categories helpful, harmless, and honest (HHH). The latter includes accuracy. Helpfulness only covers following instructions
by cubefox 4y ago
Anthropic scores fine-tuning in the categories helpful, harmless, and honest (HHH). The latter includes accuracy. Helpfulness only covers following instructions. If the model "over-censors", it declines too many requests and becomes less helpful. They say Constitutional AI is a good way to maximize both harmlessness (which is usually maximized via RLHF) and helpfulness, but not honesty. It's in the paper.
- travisjungroth 4y agoOh, thank you.