Marco Roberti

🖥 NLP Research Engineer @TechAtBloomberg, London.
👨‍🎓 AI PhD @DiUnito.

{Opinions are my own ᐸGOᐳ}

#NLP #NLProc #DeepLearning #AI

[ACL 2026 @aclmeeting]

ACL announcement:
"The ACL Executive Committee has voted to significantly change ACL's approach to protecting anonymous peer review. The change is effective immediately. (1/4) #NLPRoc

[Djamé.. @zehavoc]

Note to future #ACL2023NLP party organisers: please don't use H-index to decide who gets it or not in your party. Not only it's lame but you're just filtering out all those promising students who'll give anything for a chance to meet u and have mb an opportunity for an internship

[Marco Roberti @MarcoRoberti]

@marcostranisci @vivi_patti @tommaso_caselli @diunito Congrats, to you and to the whole group! 😎

[EMNLP 2026 @emnlpmeeting]

For a more transparent reviewing process, reviews, author responses and meta reviews will be made publicly available after acceptance notifications for the main conference accepted papers as well as opt-in rejected papers. More details: https://t.co/ie96kT6p38
#EMNLP2023 #NLProc

[Marco Roberti @MarcoRoberti]

In the last few weeks I have realized that, rather than a #moratorium on the development of AI, we *way more* need a moratorium on #journalism about #AI 🤷‍♂️

#NLP #NLProc #GPT #ML #ChatGPT #GPT4

[Marco Roberti @MarcoRoberti]

Researchers working at @TechAtBloomberg publish 4 NLP Research Papers at #EACL2023 #ICLR2023!

Come and find us (and the Terminal 🖥️) at the #Bloomberg booth in the #Dubrovnik venue during these @eaclmeeting days 🌊

#ACL #AI #ML #NLP #BloombergGPT

https://t.co/Ml81Tkoo1S

[Marco Roberti @MarcoRoberti]

Did you know? A generative system / Large Language Model, like #GPT4, #BloombergGpt, or #ChatGPT, is...

...not a database
...not an Information Retrieval system
...not linked to a knowledge base
...not a search engine

#justsaying #pedantic 😇

#GPT #GPT3 #LLM #AI #NLP

[Marco Roberti @MarcoRoberti]

Introducing #BloombergGPT, #Bloomberg’s 50-billion parameter large #language model, purpose-built from scratch for #finance.

#LanguageModel #LLM #LM #GPT #NLP #NLProc

https://t.co/vWvFcg0sgl

[Tech At Bloomberg @TechAtBloomberg]

Through a partnership between our #OSPO and Corporate #Philanthropy teams, we’ve launched @Bloomberg’s new #FOSS Contributor Fund to support the people and communities who maintain #opensource projects!
https://t.co/sgXQwzEHpa
#ImpactInAction
(1/4)

[Nate Raw @_nateraw]

Gonna start a new startup called “ClosedAI” where we just open source stuff 🫡

[Arthur Douillard @Ar_Douillard]

🚨 I'm excited about the release of 🏝️ NEVIS'22, a benchmark where we collected 106 datasets from the last 30 years of CV research!

🤖 Can you design a model to efficiently learn them all using forward transfer?

📜 My first paper while at @DeepMind 🥰

[Laure Soulier @LaureSoulier]

In case you missed my #ECMLPKDD2022 presentation on "Controlling Hallucinations at Word Level in Data-to-Text Generation"
w/ @ClementRebuffel, @MarcoRoberti, G. Scoutheeten, R. Cancelliere & P. Gallinari

here are the slides: https://t.co/tCwiaGqsfL

[Laure Soulier @LaureSoulier]

#ECML2022 - I will present tomorrow our DMDK journal paper

"Controlling Hallucinations at Word Level in Data-to-Text Generation"
with @ClementRebuffel, @MarcoRoberti, G. Scoutheeten, R. Cancelliere & P. Gallinari

PDF: https://t.co/oe87wnPAbY
code: https://t.co/0WV5iTDizu

[Marco Roberti @MarcoRoberti]

Finally, my turn to announce a #newjob!

It starts next week.
It is an #AI #Research Engineer position.
It is located in the inspiring #London headquarters of @Bloomberg.

...and it will be amazing! ✨

#Bloomberg #NLP @TechAtBloomberg

[Marco Roberti @MarcoRoberti]

@EnzoCare2 @CarloCalenda @matteosalvinimi @Calend_AI Ogni tanto ;)

[BigScience Large Model Training @BigScienceLLM]

▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓ 100%

[Marco Roberti @MarcoRoberti]

@guiodic! Mi ricordo di te dai tempi della tua appassionata (😉) partecipazione al forum #Ubuntu Italia, correva l'anno 2008 o giù di lì. Onorato per i retweet al "mio" @Calend_AI! :)

[Marco Roberti @MarcoRoberti]

Apparently, President Mattarella will be re-elected in the next ballot! The test set is being generated by the Parliament right now 🤪
In the meanwhile, this is the plot of the ballot number vs number of preferences (blue points are the model's predictions).
Have fun!
[4/4]

[Marco Roberti @MarcoRoberti]

#AI #ML #MachineLearning #Mattarella #Quirinale

On these days, the Italian Parliament is voting for the new President of the Italian Republic. A candidate is elected if he/she gets at least 505 votes.
[1/4]

[Marco Roberti @MarcoRoberti]

I trained a simple linear regression model to predict such number, based on the ordinal number of ballots, voters, and voting parties. Here are its predictions:
7^ ballot: 541
8^ ballot: 591
9^ ballot: 641
10^ ballot: 691
[3/4]