Direct Nash Optimization: Teaching language models to self-improve with general preferences
1 894
7.9
Microsoft Research334 тыс
Следующее
79 дней – 1 1676:36Популярные
11.11.23 – 5731:03:57
14.12.22 – 5181:36:23Опубликовано 3 сентября 2024, 18:59
Corby Rosset, Senior Researcher, Microsoft Research AI Frontiers, discusses teaching language models to self-improve using a preference oracle like GPT-4, framing it as a two-player game to find an optimal policy at a Nash equilibrium, and achieving state-of-the-art win rates against GPT-4 Turbo on benchmarks such as Alpaca-Eval and MT-Bench.
Microsoft Research Forum, September 3, 2024
See more at aka.ms/ResearchForum-Sep2024
Microsoft Research Forum, September 3, 2024
See more at aka.ms/ResearchForum-Sep2024
Свежие видео
3 часа – 914:09
1 день – 31 51814:28
2 дня – 40 7721:16:20
Случайные видео
333 дня – 70 5429:25
352 дня – 4393:46
17.10.23 – 62 23518:42
29.04.23 – 1 970 8713:41:36
31.12.21 – 1812:20
23 часа – 5260:49
2 дня – 289 0429:00
5 дней – 365 4369:09
7 дней – 1 94238:10
9 дней – 150 19813:43
9 дней – 6 7044:44