
Twitter account of ALMAnaCH, the Inria Paris NLP research team. @inriaparisnlp.bsky.social
Joined May 2019
- Tweets193
- Following204
- Followers1.6K
- Likes37
Inria Paris NLP (ALMAnaCH team) retweeted
🧵How to combine code correctness and efficiency objectives in online RL?
-> we break the correctness-efficiency Pareto frontier!
+125% relative improvement in CWM32B compared to standard RLVR (13.7 ->30.9 pass@1@top30%)
+150% with Qwen32B
Paper: arxiv.org/abs/2607.25970
arxiv.org/abs/2607.25970
Exploration of RLVR setups for code optimization.
Information about ALMAnaCH’ seminar today is available on the alternative platform we just migrated to (as well as on our web site).
The ALMAnaCH project-team at Inria Paris is now on an alternative platform (same username, i.e. inriaparisnlp)! Follow us there for news and updates about our research and seminar announcements!
Congratulations to Nathan Godey (@nthngdy) who successfully defended his PhD on the 20th December entitled “Improving Representations for Language Modeling” supervised by Benoît Sagot (@bensagot) and co-supervised by Éric de La Clergerie (@DeVillemonte)!
We are pleased to announce our next seminar by Caio Corro (@INSA_Rennes) entitled "Named-Entity Recognition: Resurrecting Old School Machine Learning in
the Era of Deep Learning" on Friday 20/12/2024 at 11am CET. Details➡️almanach.inria.fr/seminars-e…
Connection link here➡️ docs.google.com/document/d/1…
Congratulations to Floriane Chiffoleau @MissBrutus who successfully defended her PhD last Wed. “Comprendre le processus de reconnaissance automatique de texte: entraînement de modèles, vérité de terrain et erreurs de prédiction” sup. by @AnneBaillot (@LeMansUniv) & @laurentromary
Congratulations to @CyrilChhun who successfully defended his PhD last Tuesday on a “Meta-Evaluation Methodology and Benchmark for Automatic Story Generation” supervised by Fabian Suchanek (@telecomparis) and @ChloeDClavel!
We are pleased to announce our next seminar by @RaphaelBaena (@ImagineEnpc, École des Ponts ParisTech) entitled "A General Framework for Text Line Detection
and Recognition" on Friday 15/11/2024 at 11am CET. Details➡️almanach.inria.fr/seminars-e…
Connection link here➡️ webinaire.numerique.gouv.fr/…
Inria Paris NLP (ALMAnaCH team) retweeted
I am a bit late to the party but I am happy to announce that our paper "Tree of Problems: Improving structured problem solving with compositionality'" was accepted to the #EMNLP2024 main Conference 🤩🔥.
So, what is Tree🌴 of Problems (ToP)?
1/8
We are pleased to announce ou next seminar by @MarineCarpuat (University of Maryland, currently on sabbatical at @InriaParisNLP) entitled "Beyond Translation: Human-Centered NLP for Cross-Lingual Communication" on Friday 11/10/2024 at 11am CEST. Details➡️almanach.inria.fr/seminars-e…
Connection link here➡️ docs.google.com/document/d/1…
Inria Paris NLP (ALMAnaCH team) retweeted
📣 Crowdsourcing Call !
📚 You have books ?
📱You are reading this on your phone ?
📸 Reply to this tweet with 1 or more picture of a page where we see one of your finger, your palm, etc.
📖 We are creating a dataset based on real world situation for segmenting docs.
Example:
We are excited to announce our next seminar, co-organised w/ @ISIR_labo by Paolo Rosso (@UPV @fvalgrai): “Beyond the detection of fake news and explicit hate speech: conspiracy theories and implicit hate speech with stereotypes, jokes and sarcasm”, Thurs 4/7, 11am CEST.
Details➡️almanach.inria.fr/seminars-e…. Connection link to be shared on the day!
Connection link here➡️docs.google.com/document/d/1…
We are thrilled to announce our next seminar, co-organised w/ @ISIR_labo. Nicolas Rollet (@telecomparis) will be talking about “Interaction Humain-Machine, IA parlante et ethnométhodologie : le diable est dans les détails”, Friday 21/6, 11am CEST. Details➡️almanach.inria.fr/seminars-e…
Connection link here ➡️ docs.google.com/document/d/1…
Inria Paris NLP (ALMAnaCH team) retweeted
Announcing mOSCAR, multilingual interleaved text-image corpus as part of @oscarnlp project.
Paper: arxiv.org/pdf/2406.08707
Dataset: huggingface.co/datasets/osca…
Doc: oscar-project.github.io/docu…
1/6


















