
How RLHF Makes AI More Effective and Reliable
This is the CitePep accuracy record for “How RLHF Makes AI More Effective and Reliable”, published by When Notes Fly. It lists every change experts checked and accepted on that page.
What RLHF is, a 1.3 billion parameter model trained this way beat GPT-3 at 175 billion in human preference ratings.
What has been corrected on this page?
Every accepted correction to this page is recorded with the exact change, so readers can see how the page improved over time.
-
A claim-level statistics check of this article: 1 claim that did not match the published source was corrected to what the source reports.
BeforeA 2023 study by Anthropic found that inter-annotator agreement on preference comparisons ranged from approximately 63% to 77% depending on the difficulty of the comparison - meaning that on the hardest comparisons, labelers disagreed nearly as often as they agreed.
AfterAnthropic reported an average agreement rate of only about 63% between its own researchers and its crowdworkers on preference comparisons.
Why: Each study and survey figure in the article was checked against the published source. Figures that differed from the source were corrected to the source's own numbers, dates and attribution.
View the full record →
Who checked this page?
1 contributor has checked "How RLHF Makes AI More Effective and Reliable" on When Notes Fly. Each name below links to that person's public CitePep profile, where every contribution they have made is listed with the exact change they proposed.