Maja Trebacz
@majatrebacz
Followers
987
Following
249
Media
0
Statuses
13
Explore trending content on Musk Viewer
#ใจซ่อนรักEP5
• 710863 Tweets
LINGORM TSOU EP5
• 700347 Tweets
Cheatle
• 423334 Tweets
#ブルアカ
• 265002 Tweets
Secret Service Director
• 210262 Tweets
सुप्रीम कोर्ट
• 141610 Tweets
ログイン
• 65831 Tweets
Jamaican
• 64269 Tweets
Nancy Mace
• 52858 Tweets
Roro
• 52444 Tweets
travis
• 51594 Tweets
#KingOfHeartsJaehyukDay
• 45432 Tweets
Our Shining Star Yoon Jaehyuk
• 43806 Tweets
Rajoy
• 43503 Tweets
#いつもしあわせでいてくれたらありがとう
• 40009 Tweets
Jim Jordan
• 34817 Tweets
Raskin
• 33435 Tweets
Kris Tyson
• 30606 Tweets
シロコテラー
• 30459 Tweets
谷川俊太郎
• 20784 Tweets
Spawn
• 16067 Tweets
Çanakkale
• 15643 Tweets
臨戦ホシノ
• 14871 Tweets
Bodrum
• 12630 Tweets
灼熱カバディ
• 11649 Tweets
Chalobah
• 10091 Tweets
Btw, I joined
@OpenAI
and this is what I’ve been up to so far.
We've just released a paper on training LLM critics to enhance human feedback for training LLMs.
Kudos to the incredible team —
@nmca
,
@gadzin1203
,
@agentydragon
, Juan,
@janleike
Excited for what's ahead 🚀
We’ve trained a model, CriticGPT, to catch bugs in GPT-4’s code. We’re starting to integrate such models into our RLHF alignment pipeline to help humans supervise AI on difficult tasks:
603
2K
7K
16
30
603
𓅪𓅪𓅪 Sparrow 𓅪𓅪𓅪
It was amazing to work on this dialogue agent and train it from human feedback! Sparrow searches Google to improve and back up its claims, and follows a set of rules to be less harmful.
Large language models can exhibit falsehoods, discriminatory language, and other unsafe behaviour. Introducing Sparrow: a dialogue agent that can search the internet and is trained to be more helpful, correct, and harmless using RL from human feedback: 1/
17
201
864
0
16
174
We tuned a massive language model to support its answers with quotes from the web.
I'm delighted to share this after putting much effort into it. Privileged to work on this project at
@deepmind
and collaborate with an amazing team
@jacobmenick
,
@__nmca__
,
@geoffreyirving
et al.
Introducing GopherCite, a fine-tuned version of Gopher that used human feedback to learn to back up claims with supporting evidence from the web. GopherCite can also answer questions about a given document or abstain when unsure.
Learn more: 1/
9
137
619
0
6
43
@jacobmenick
Thank you, and right back at you!
It was a great pleasure and learnt a ton from you :D
0
0
2