Camille Couturier

Applied researcher @Microsoft; ideas my own

[Mehrdad Farajtabar @MFarajtabar]

1/ Can Large Language Models (LLMs) truly reason? Or are they just sophisticated pattern matchers? In our latest preprint, we explore this key question through a large-scale study of both open-source like Llama, Phi, Gemma, and Mistral and leading closed models, including the recent OpenAI GPT-4o and o1-series.
https://t.co/2tv8Pp9MSz

Work done with @i_mirzadeh, @KeivanAlizadeh2, Hooman Shahrokhi, Samy Bengio, @OncelTuzel.

#LLM #Reasoning #Mathematics #AGI #Research #Apple

[Camille Couturier @couturierc]

Le CO2 réellement "sauvé" est celui que l'on n'émet pas.
Mon prochain déplacement longue distance, même si effectué en train, est-il réellement utile et nécessaire ? 🙃🤔

@GroupeSNCF @SNCFConnect @SNCFVoyageurs #TeamTrain

[Camille Couturier @couturierc]

Je suis végétarien, notamment pour des raisons d'empreinte écologique, et je ne trouve pas cette comparaison drôle ou appropriée.
@GroupeSNCF @SNCFConnect @SNCFVoyageurs #TeamTrain

[Denis Wirtz @deniswirtz]

Here is our largest database of PhD fellowships yet: we found 195 fellowships and funding opportunities.

For each fellowship, this searchable database lists description, deadline, amount, and eligibility criteria.

Download it here: https://t.co/1h76jJaHOv

[Camille Couturier @couturierc]

@DrJimFan Something to integrate in @myplume?

[Camille Couturier @couturierc]

@MrsCaroline_C @ingridtherwath @MaterniteBluets Félicitations @MrsCaroline_C et @ingridtherwath ! 🎉👶

[Alon Albalak @AlbalakAlon]

TL4NLP workshop is now online!
https://t.co/ObpkaflSIB

Thread for speaker times:
Jonas Pfeiffer @PfeiffJo
Graham Neubig @gneubig
Percy Liang & Ananya Kumar @percyliang @ananyaku
Sara Hooker & Kyunghyun Cho @sarahookr @kchonyc
David Adelani @davlanade
Mike Lewis @ml_perception
👇

[Camille Couturier @couturierc]

@SFR Merci de bien vouloir supprimer mon adresse email, utilisée sans mon consentement, de vos bases de données.

[Ananya Kumar @ananyaku]

How should you fine-tune a large pretrained model (CLIP, SimCLR) robustly? We find that standard fine-tuning can do poorly out-of-distribution (test data ≠ fine-tuning data). Our analysis leads to a simple fix, higher accuracy on 10 datasets. https://t.co/hyfx9vnAOd (ICLR Oral)

[Alfredo Canziani @alfcnz]

An awesome @geoffreyhinton presenting «The Forward-Forward Algorithm for Training Deep Neural Networks».
A training algorithm that does not require to propagate derivatives or remember neural activities and learns representations of streaming sensory data on the fly. #NeurIPS22

[Luiz Pessoa @PessoaBrain]

𝗠𝘂𝗹𝘁𝗶𝘀𝗰𝗮𝗹𝗲 𝗯𝗿𝗮𝗶𝗻 𝗺𝗼𝗱𝗲𝗹𝗶𝗻𝗴 as the way to go!
"The quest for multiscale brain modeling", by Egidio D’Angelo and Viktor Jirsa.
How do all scales influence each other??
https://t.co/CMDectjyVw

[Camille Couturier @couturierc]

@thalys_en It seem that the https://t.co/BcsEXz6Dml website is down (?)

[Camille Couturier @couturierc]

@ParisAeroport Bonjour, comment puis-je mettre un objet dans la consigne Tripperty au terminal 2E?

[Camille Couturier @couturierc]

@TERPays2LaLoire Bonjour, dans le TER 857323 entre Le Mans et Nantes. Il fait super froid (clim à fond). Y at-il possibilité de moduler la clim ? 2 enjeux: 1/ climatique et économie d'énergie 2/ ne pas attraper de rhume en plein juillet. :-)

[Brian Lester @blester125]

My first @GoogleAI residency project was accepted to @emnlpmeeting #EMNLP2021!

Prompt Tuning can condition a frozen T5 XXL model to perform new tasks while only adding 0.003% more parameters and no performance loss.

Camera Ready 📸: https://t.co/UMSpXpidmH

Quick Thread 🧵(1/7)

[Akari Asai @AkariAsai]

Introducing 𝗔𝗧𝗧𝗘𝗠𝗣𝗧, a new modular, multi-task, and parameter-efficient approach to combine knowledge from multiple tasks to solve a new task using a small trainable of parameters 🔥 while keeping the original LM *frozen* 🧊 [1/9]

Paper 📜: https://t.co/IgJvwCX5tU

[Camille Couturier @couturierc]

@wiseed Bonjour, le projet fonds France Valley Revenu Europe mentionne 10% de commission de souscription. A qui est reversée cette commission et que finance-t-elle?
https://t.co/PlwnPaYJoQ

[Katherine Lee is at ICML! @katherine1ee]

Language models memorize data & we can pull that data back out.

If you're training on private data, this should give you pause. If you're training on public data, this should still give you pause. Where does your training data come from and who consented (or didn't)?

[Thomas Zimmermann @tomzimmermann]

The Productivity + Intelligence group in @MSFTResearch seeks experienced research managers in empirical software engineering to join us in advancing research on data-driven and AI-infused developer productivity experiences. For more information: https://t.co/ccxyHcgsHy

[Michele Tufano @tufanomichele]

Check out our work at Microsoft on Unit Test Case Generation with Transformers!

Models pretrained on English and source code, then finetuned on the task of generating test cases.

Paper: https://t.co/QYpcpXYV9I

Collaborators: @DawnDrainML, @asvyatko, @skdeng, Neel Sundaresan