3 ms·
This could work really well if it replaced GPT-X-judged performance ranking with human-in-the-loop ranking of prompts, but that’s not as exciting, I guess.
by namuol 3y ago
This could work really well if it replaced GPT-X-judged performance ranking with human-in-the-loop ranking of prompts, but that’s not as exciting, I guess.
- __loam 3y agoI think the human in the loop evaluation of prompts is a red herring since the LLMs were trained with HFRL, which optimized them to be convincing. Humans aren't reliable for evaluating these things because it was literally trained to make us think it's doing well.
- namuol 3y agoI can’t see how we’d expect the same model to be any better…