AI & ML
Fine-tuning GPT-2 from human preferences
OpenAI Blog0 views
We’ve fine-tuned the 774M parameter GPT-2 language model using human feedback for various tasks, successfully matching the preferences of the external human labelers, though those preferences did not always match our own. Specifically, for summarization tasks the labelers preferred sentences copied wholesale from the input (we’d only asked them to ensure accuracy), so our models learned to copy. Summarization required 60k human labels; simpler tasks which continue text in various styles required only 5k. Our motivation is to move safety techniques closer to the general task of “machines talking to humans,” which we believe is key to extracting information about human values.
Source:https://openai.com/index/fine-tuning-gpt-2Related posts

Active Debt vs Settled Debt: The Lifecycle State Machine Inside FinLedger's Lending Engine
0dev.to UX
Automatic clock-in for construction crews: designing for bad GPS
0dev.to UX
10 Food Puns That Will Make You Smile at the Dinner Table
0dev.to Watercooler

I Upgraded to a Portable OLED Monitor for Remote Coding & Gaming—Here is What Changed
0dev.to Watercooler
Mauricio Pochettino và bài toán tái thiết cấu trúc phòng ngự Man Utd trước thềm đại chiến Liverpool
0dev.to Data
Crystal Palace bổ nhiệm Thomas Frank: Bài toán tái thiết từ đáy bảng xếp hạng Premier League
0dev.to Data