Hoy, 25 de septiembre, se cumplen 102 años desde que el abuelo de Encarnita iba arando con los mulos y se topó con una piedra que resultó ser un Batipsterio Romano del Siglo I d.C. ¿A quién no le va a gustar?

❤ Jesús Quintero eterno. D.E.P.

Si soy, no me escondo

Demà tenim databeersvlc.com amb ponents de luxe per a xarrar d'una València +pròxima, +intel·ligent, +sostenible 😍 Apunta't a conèixer Fernando Gallego Fernando Gallego , Aida Villalba, Ernesto Faubel Ernesto Faubel y Carol Soriano

SPOILER

Molt preparat VOX per a governar, la veritat.

[1/7] Introducing "Investigating the translation capabilities of Large Language Models trained on parallel data only" To our knowledge, the first work studying translation on #LLM trained exclusively on parallel data. Arxiv paper: arxiv.org/abs/2406.09140

#NewProfilePic

📢 Introducing MixerMDM [#CVPR2025] — the first learnable model composition technique for combining pre-trained human motion diffusion models while preserving their core characteristics. 🟡➕🔵🟰🟢 MixerMDM enables consistent, fine-grained individual controllability in

🚨New Paper🚨 1/ We often assume that well-written text is easier to translate ✏️ But can #LLMs automatically rewrite inputs to improve machine translation? 🌎 Here's what we found 🧵

Let's chat at #NAACL2025 about evaluation et al! ⚗️

Multilinguality is happening at #NAACL2025 Xinyu Crystina Zhang Nathan Brown @ NAACL 2025 Dayeon (Zoey) Ki Rena Gao Ona de Gibert

🎉 Exciting first session with paper presentations and a keynote talk at the AmericasNLP 2025 workshop! We have much more lined up for the rest of the day -- join us in-person or on Zoom! ✨ #NLProc #NAACL2025

1/ How can a monolingual English speaker 🇺🇸 decide if a French translation 🇫🇷 is good enough to be shared? Introducing ❓AskQE❓, an #LLM-based Question Generation + Answering framework that detects critical MT errors and provides actionable feedback 🗣️ #ACL2025

Prev. work by Goodfire, neuronpedia & al. showed that SAEs can be helpful beyond interpretability, to surgically condition model behaviors. In this work, we take a small step towards making steering useful for MT personalization on literary works! Check out our 🧵⬇️

Excited to spend the week at #CVPR2025 in Nashville! 🇺🇸 If you’re interested in learning more about MixerMDM, come check out our poster at the following sessions: 📍 💃 [HuMoGen - CVPR Workshop 2025] Wednesday 11, 16:00-18:00, ExHall D #430 📹 [AI4CC] Thursday 12, 17:00, ExHall D #412 👑

¿A qué edad te enteraste que Google tiene una herramienta GRATUITA para presentar datos de forma más dinámica, fácil y llamativa?📊🤯 Se llama Data Gif Maker y está disponible aquí→ datagifmaker.withgoogle.com

I've written the full story of Attention Sinks — a technical deep-dive into how the mechanism was developed and how our research ended up being used in OpenAI's new OSS models. For those interested in the details: hanlab.mit.edu/blog/streaming…

Induction heads are surprisingly powerful. In a new preprint, we find that they can learn what to attend to in-context! We study this in a hierarchical prediction task and uncover a possible mechanism giving rise to in-context learning in induction heads. See thread for details!

To read more, check out our new paper, out now: Bayesian Influence Functions for Hessian-Free Data Attribution arxiv.org/abs/2509.26544 Philipp A Kreer, Wilson Wu, Max, Zach Furman, Jesse Hoogland 10/10