Re:Compare how GPT-2, 3, 3.5 and 4 answer the same questions
by DrawTR on Tuesday October 31, 2023 @04:40PM
(#38091069)
For what it's worth, not all of these examples are consistent: https://chat.openai.com/share/29e1c2bd-ef7b-4475-b5a9-9287d1...
[ Parent
| on HN ]
Re:Compare how GPT-2, 3, 3.5 and 4 answer the same questions
by jellyberg on Tuesday October 31, 2023 @05:47PM
(#38091865)
Yeah - worth noting that we use temperature=0 for reproducibility while ChatGPT I think uses t=0.7. We also prefix the prompt with few-shot examples of questions and answers with chain of thought examples to elicit the models' full capabilities.