tiptales retweeted
magnet:?xt=urn:btih:5546272da9065eddeb6fcd7ffddeef5b75be79a7&dn=mixtral-8x7b-32kseqlen&tr=udp%3A%2F%2Fopentracker.i2p.rocks%3A6969%2Fannounce&tr=http%3A%2F%https://xitter.cf/t.co/g0m9cEUz0T%3A80%2Fannounce
RELEASE a6bbd9affe0c2725c1b7410d66833e24
tiptales retweeted
Here's an incomplete list of things we never talk about when we talk about LLMs and AI vickiboykis.com/2023/07/18/w…
tiptales retweeted
Excited to share that my internship project was accepted as a long paper in NAACL 2022! In “TRUE: Re-evaluating Factual Consistency Evaluation”, we present a comprehensive study of factual consistency metrics to help mitigate hallucinations in text generation.🧵1/7 @naaclmeeting
tiptales retweeted
« Fake News et crédibilité de l’information publique : le regard des technologies de la langue » : thématique de #TEDonnee2022, journée d’étude organisée par les doctorants du @CentreTesniere, qui se tiendra le 01/04 à la @MSHE_Ledoux ! Plus d’info ici: tinyurl.com/4n4v2wtc
tiptales retweeted
You've probably seen results showing impressive few-shot performance of very large language models (LLMs). Do those results mean that LLMs can reason? Well, maybe, but maybe not. Few-shot performance is highly correlated with pretraining term frequency. arxiv.org/abs/2202.07206
tiptales retweeted
🎉Great news: the winning teams of our FNP shared tasks will receive a free registration to attend @lrec2022 🇫🇷 to present their work. The free registrations are provided by ELRA elra.info who are generously supporting FNP workshop this year. wp.lancs.ac.uk/cfie/fnp2022/
tiptales retweeted
Free full-day workshop materials on Graph Neural Networks @StanfordEng are available!
The workshop brings together leaders from academia & industry.
Find recordings of talks and slides here: snap.stanford.edu/graphlearn…
tiptales retweeted
Le forum en ligne "Innovation, technologies et plurilinguisme" va s'ouvrir. A noter demain 14h une session #APIL sur le thème du NLP industriel. Venez nombreux apil-asso.fr/presidence-fran… cc @apil_asso
tiptales retweeted
🎉Our “Imagine before You Speak” paper got into @iclr_conf. The idea of using scene structures to guide commonsense-based text generation is neat. This work gets to pull a range of resources for creating a scene imagination module and show the great potential.
🤔Imagine before you 🗣️describe a scene! Check out our new preprint on the generative commonsense reasoning task, which is to compose common concepts into commonsensical descriptions using natural language. [1/n]
arxiv.org/abs/2112.06318
tiptales retweeted
"Representing Long-Range Context for Graph
Neural Networks with Global Attention" by @parasjain @Michaelvll1 et al. at #NeurIPS2021
Strong results on graph classification when combining GNN --> Transformers.
PDF: openreview.net/forum?id=nYz2…
tiptales retweeted
In prep for #NeurIPS2021 next week, I've made a condensed video explaining both papers I'm presenting on 🌍World Models🌎 in text. A shiny new benchmark and state of the art KG model, the Worldformer!! Drop by during the conference to chat more!
Video🎥: youtu.be/Snv2KYKsVF4
ALT A picture of a scroll with the title "Modeling Worlds in Text by prithviraj Ammanabrolu and Mark Riedl" from the "Allen institute for AI and Georgia Tech" with the GitHub link in the next tweet
Parte the second thread, as promised:
Here's the Worldformer: a sota text game world model that multi-task learns to generate all possible lang actions and the *difference* between world states as a knowledge graph, using it to learn env dynamics!
Paper: arxiv.org/abs/2106.09608
ALT Worldformer architecture diagram showing the multi task architecture
ALT Difference in world states as knowledge graphs between subsequent states
tiptales retweeted
Our new state-of-the-art algorithm NBFNet for knowledge graph reasoning (to appear @NeurIPSConf '21) is open sourced: github.com/DeepGraphLearning….
NBFNet is a general GNN framework for link prediction, and works in both transductive and inductive setting, by @zhu_zhaocheng @Oxer22
Code of NBFNet is released at github.com/DeepGraphLearning…. Check it out!
tiptales retweeted
N-grammer: Augmenting Transformers with latent n-grams
Substantially outperforms Transformer and Primer on C4 LM by augmenting the model with n-grams constructed from a discrete latent representation of the text sequence.
openreview.net/forum?id=GxjC…
tiptales retweeted
Rappel : rejoignez @mazancourt de Mazancourt Conseil pour notre prochain meetup @ElasticFR demain à 18h à propos de "#Elasticsearch pour analyser les discours politiques en ligne" ! Inscriptions 👉🏼 community.elastic.co/e/myvcn…
tiptales retweeted
The recordings of our FNP 2021 workshop are now available on YouTube and our website. This includes all the keynote speakers, paper presentations as well as the FNP panel discussion. youtube.com/playlist?list=PL… wp.lancs.ac.uk/cfie/fnp2021/
tiptales retweeted
We would like to thank everyone who contributed to the success of our FNP2021 workshop. This year our speakers & attendees came from 87 cities from all around the world. Thanks & look forward to seeing you again in future events 😊 @dsilancaster @UCREL_NLP wp.lancs.ac.uk/cfie/fnp2021/ – at North West, England
tiptales retweeted
Ooops did you just open source GPT-3 ??
You can now upload models, datasets, and spaces directly through the @huggingface hub interface so let me try with a small model I had locally 😊😊😊