Caio

« On veut récupérer le Mama Shelter pour le transformer en centre social d’urgence » Transformer un hôtel de luxe en centre social d’aide d’urgence pour les personnes sans-abri, les personnes LGBT, les femmes victimes de violences : ce qu’on défend à Rennes avec RP 👇

Je supporte plus vivre dans une société qui laisse des gens MOURIR de FROID dans la rue alors qu’il y a des logements vides partout en France. On ne peut pas s’habituer à ça.

seven 200-line additional result tables just convinces me that you need to rewrite the paper, not that the review is wrong. Better to remain silent or just respond simply rather than try to Prove! Them! Wrong!

🇩🇪🪖 Grève des élèves contre le retour du service militaire ! Ce vendredi en Allemagne, 50.000 collégiens et lycéens se sont mis en grève dans plus de 90 villes contre la réintroduction du service militaire. Alors que le gouvernement mène une campagne agressive dans les écoles et l'espace public pour faire la promotion de l'armée, les jeunes rappellent avec force qu'ils refusent de servir de chair à canon !

Rennes 2 la Rouge, on se retrouve ce soir à 18h pour discuter de la situation politique explosive. Aller à Rennes 2 un jour de chute du gouvernement… la révolution est proche 😂✊🏼🇵🇸

5) EuroBERT: Scaling Multilingual Encoders for European Languages w/ @N1colAIs @gisship @DuarteMRAlves @AyoubHammal @UndefBehavior @Fannyjrd_ @ManuelFaysse @peyrardMax @psanfernandes @RicardoRei7 @PierreColombo6 @tomaarsen - Poster session 5, Thu Oct 9, 11:00 AM – 1:00 PM

@vaiter Self-promotion: we recently applied this for parallelizable inference in linear-chain CRF (i.e. instead of using Viterbi/forward algorithms), and it works quite well! :) arxiv.org/abs/2506.00732

Cyclic projection onto convex sets is a method to find a point in the intersection of convex sets. It iteratively projects onto each set in a cycle, and it converges to a point in the intersection under restrictive assumptions.

As you react to EMNLP rebuttals, Please make sure to substantiate your stance on why the score you have chosen is justified Keeping the information latent doesn't help authors understand your point to improve their work, and doesn't help AC/SAC make a grounded and fair decision.

Simple example for how errors can creep into our papers through LLM use: I had a statistic for 2020. I googled the same statistic for 2023. AI overview tells me the statistic for 2023 & provides a link to support the claim. The link is to a 2023 article citing the 2020 statistic.

Moreover, we show how to learn parameters of a Bregman CRF in supervised and partially supervised settings using (partial) Fenchel-Young losses. Link to the paper: caio-corro.fr/pdf/bregman_cr…

Forget about Viterbi and forward algorithms! In this paper ( #acl2025 #ACL2025NLP ), we introduce Bregman CRFs for sequence labeling. We propose a novel (approximate) inference algorithm based on iterative Bregman projections which can take full advantage of modern GPUs.

Having trouble choosing a track for your ARR / EMNLP submissions? Take a look at this blog post for more information: 2025.emnlp.org/track-changes/

If you submitted to ICML and TMLR as an author, where did you get higher review quality (grounded, fair, and helped improve your paper)? 𝐏𝐥𝐞𝐚𝐬𝐞 𝐝𝐞𝐜𝐨𝐮𝐩𝐥𝐞 𝐭𝐡𝐞 𝐮𝐥𝐭𝐢𝐦𝐚𝐭𝐞 𝐟𝐚𝐢𝐭𝐡 𝐨𝐟 𝐩𝐚𝐩𝐞𝐫 𝐟𝐫𝐨𝐦 𝐲𝐨𝐮𝐫 𝐣𝐮𝐝𝐠𝐞𝐦𝐞𝐧𝐭 𝐨𝐟 𝐪𝐮𝐚𝐥𝐢𝐭𝐲.

Current KL estimation practices in RLHF can generate high variance and even negative values! We propose a provably better estimator that only takes a few lines of code to implement.🧵👇 w/ @xtimv and Ryan Cotterell code: arxiv.org/pdf/2504.10637 paper: github.com/rycolab/kl-rb

ARR May (EMNLP) cycle opens May 5th. Reminder: (1) All resubmissions from past cycles MUST be declared, or they will be desk-rejected (automated checks against past cycles apply); (2) The responsible checklist must be carefully completed, or the paper gets a desk reject.

Today I had my paper rejected by ICML. I don’t like to complain about the free voluntary work that the AC and reviewers do, but this was one of the most careless reviews and AC we ve had. Our rejection was based on this:

On a un souci avec la sociologie des étudiants en médecine. Pour pouvoir se balader avec une pancarte pareille sans que tes camarades te disent « mec t’es au courant que Mulhouse c’est une super grande ville ? » (et pas vraiment un désert médical en plus) faut un sacré entre soi

ModernBERT or DeBERTaV3? What's driving performance: architecture or data? To find out we pretrained ModernBERT on the same dataset as CamemBERTaV2 (a DeBERTaV3 model) to isolate architecture effects. Here are our findings:

Bah la team pk vous masquez mon commentaire ? ils sont public nos résultats pourtant ^^ (À gauche sur mon compte et à droite sur celui d'un pote)