Pinned Tweet
I am happy to share that together with @kilian_maciej, @rom1504, @wightmanr and the rest of the github.com/mlfoundations/ope… community we added a new model architecture to the repository, Contrastive Captioners (CoCa), for all the info check laion.ai/blog/coca/.
Briefly 👇
giovanni retweeted
Fact-Checking the Output of Large Language Models via Token-Level Uncertainty Quantification aclanthology.org/2024.findin… now at poster 95 #ACL2024 #ACL #NLProc
giovanni retweeted
🎉SAC Awards:
10) Speech Translation with Speech Foundation Models and Large Language Models: What is There and What is Missing? by Gaido et al.
11) AI 'News' Content Farms Are Easy to Make and Hard to Detect: A Case Study in Italian by Puccetti et al.
#NLProc #ACL2024NLP
giovanni retweeted
v2.23.0 of OpenCLIP was pushed out the door! Biggest update in a while, focused on supporting SigLIP and CLIPA-v2 models and weights. Thanks @gabriel_ilharco @gpuccetti92 @rom1504 for help on the release, and @bryant1410 for catching issues. There's a leaderboard csv now!
There is code to train and evaluate CoCa in OpenCLIP github.com/mlfoundations/ope…. There are also two pretrained versions, coca_ViT-B-32 and coca_ViT-L-14 and their fine-tuning on mscoco.
Thanks to lucidrains sigmoid.social/@lucidrains for the original implementation github.com/lucidrains/CoCa-p…
giovanni retweeted
📢 #NLPaperAlert: Outliers Dimensions that Disrupt Transformers Are Driven by Frequency
with @gpuccetti92 @bkbrd Felice Dell'Orletta
TLDR: remember how BERT over-relies on just 48 magic params? They're related to token frequency in training! /1
arxiv.org/abs/2205.11380
giovanni retweeted
New recent examples added to Javis from our community! They show:
1. Galton Board to approximate a normal distribution
2. Central limit theorem
Makes me want to brush up on my #Statistics and #Calculus !
Check out Javis: github.com/Wikunia/Javis.jl
#julialang #Opensource
(1/3)
