Re:What we learned from using GPT for 500k+ classifications
by mistrial9 on Wednesday July 12, 2023 @09:42AM
(#36694379)
amazing that this streaming pile of characters and its uncreative associations with three-letter-agency code names, results in exactly ninety two percent accuracy.. almost like its profoundly wrong in exactly the most important ways
[ Parent
| on HN ]
Re:What we learned from using GPT for 500k+ classifications
by 19h on Wednesday July 12, 2023 @02:21PM
(#36699338)
Care to elaborate? Not sure why this tone is appropriate.
The 92% is an average and not the exact accuracy across all categories; the accuracy varies by category as every category is represented by its own filter.