Eneko Agirre
banner
eagirre.bsky.social
Eneko Agirre
@eagirre.bsky.social
190 followers 180 following 20 posts
Prof. https://ehu.eus Head of HiTZ research center @hitz-zentroa.bsky.social, Spanish Informatics Research Prize 2021. ACL fellow @aclmeeting.bsky.social. Elected member https://jakiunde.eus I post in Basque, Spanish and English https://hitz.eus/eneko
Posts Media Videos Starter Packs
Egun handia 🎉

Hobetzen jarraituko dugu baina zuen laguntza ere behar dugu. Erabili eta lasai esazue 👍 edo 👎

latxa.hitz.eus
#Latxa txatbota probatarako erabilgarri jarri dugu! latxa.hitz.eus

Jaso ditugun eskaerei erantzunez zuen eskura jarri dugu Latxaren bertsio ahaltsuena, chatGPT-tik gertu dabilena, baina euskara txukunagoa sortuz.
Gradio
Click to try out the app!
latxa.hitz.eus
Reposted by Eneko Agirre
HiTZ zentroko kideek aurreko asteburuan mentore gisa laguntzen egon ziren Innovation Banking Hack Fest-ean #IBHF25. Bertan ALIA proiektuko ereduak, #Latxa barne, erabiltzen irakatsi zieten parte-hartzaileei.

www.sabadelldigital.com/events/innov...
Reposted by Eneko Agirre
Hemos estado en la reunión proyecto ILENIA con el equipo de IA, TL y modelos de lenguaje de la Dirección General de IA (Aleida Alcaide) de la SEDIA
Reposted by Eneko Agirre
Ostegun honetan, @eagirre.bsky.social-ek Gipuzkoako Foru Aldundiaren Plaza podcastean parte hartuko du beste hizlari batzuekin batera. "Adimen Artifiziala: erronkak, aukerak eta mehatxuak" gaia jorratuko dute.

Informazio gehiago: www.gipuzkoa.eus/eu/plazapodc...
Plaza podcasta - gipuzkoa
www.gipuzkoa.eus
Very interesting dataset related to a yet unsolved problem: how do children learn language. Basque and other 44 languages! Glad that our center is involved with co-author @juletxara.bsky.social
🌍Introducing BabyBabelLM: A Multilingual Benchmark of Developmentally Plausible Training Data!

LLMs learn from vastly more data than humans ever experience. BabyLM challenges this paradigm by focusing on developmentally plausible data

We extend this effort to 45 new languages!
Reposted by Eneko Agirre
We are happy to announce the next webinar in the Language Technology webinar series!

Speaker: Jose Camacho-Collados (Cardiff University)

Title: How Language Models Navigate Culture in a Multilingual World

Date: Thursday, October 16, 15:00 CET

Registration (free): www.hitz.eus/webinar_izen...
Irail hau hain izan da bizia (eta urriraino iristen ari da) sare sozialei kasu egitea ohikoa baino gehiago kostatzen ari zaidala.

Adibide bezala duela bi aste eduki genuen ekitaldi gako hau, gure @hitz-zentroa.bsky.social zentroaren gorako bilakaeran mugarria dena.

Bi asteko atzerapena 😅
Ostiralean, Adimen Artifizial eta Hizkuntza Teknologiako HiTZ Katedraren #AI_LT_catedra aurkezpena izan zen Miramar Jauregian eta Sedia, @ehu.eus, @gobeus.bsky.social, Avature, Orai NLP, Ikerlan, Multiverse, @tecnalia.bsky.social, Mas Orange, @artetxem.bsky.social lagunek parte hartu zuten.
Este septiembre está siendo tan intenso que me cuesta encontrar el tiempo para antender las redes sociales.

Como muestra este evento tan importante, todo un hito en la evolución de nuestro centro @hitz-zentroa.bsky.social
El viernes se inauguró oficialmente la Cátedra HiTZ de Inteligencia Artificial y Tecnología del Lenguaje #AI_LT_catedra en el Palacio Miramar en la que participaron Sedia, @ehu.eus, @gobeus.bsky.social, Avature, Orai NLP, Ikerlan, Multiverse @tecnalia.bsky.social, Mas Orange, @artetxem.bsky.social
Reposted by Eneko Agirre
Ayer presenté mi proyecto de tesis "Document Level Information Extraction in Low-Resource Languages" en el simposio doctoral del congreso #SEPLN2025 ¡Eskerrik asko a los organizadores!
Reposted by Eneko Agirre
Bilboko Areatzan, Ikertzaileen Europako gauean #Latxa demo batzuk erakutsi genituen #AI_LT_catedra
Highly recommended, taught by one of our top senior researchers!
Next week, the third course "Introduction to LT Applications" organised by #AI_LT_catedra will start.

Last seats available! Do not miss it!

www.hitz.eus/iltapp/
Introduction to Language Technology Applications
www.hitz.eus
En @hitz-zentroa.bsky.social estamos convencidos de que el impacto de la IA no va a dejar de crecer. Excelente noticia!
Hau da ikasturtea ondo hastea!
We are very proud of our research: we managed to build a usable open-weight LLM which is competitive with frontier models in an open LM arena, and the method to adapt Llama to Basque plus the thorough evaluation methodology have been reviewed.
Reposted by Eneko Agirre
We are thrilled to announce that the research which lead to the best open-weight LLM for Basque, has been accepted at EMNLP, one of the premier conferences in NLP and AI.
¡Así da gusto empezar el curso!
Este agosto hemos recibido una noticia excelente sobre #Latxa. ¡El artículo científico que presenta el método de construcción de Latxa ha sido aceptado en EMNLP, uno de los congresos más importantes de nuestro ámbito!
Reposted by Eneko Agirre
Oso pozik gaude! HiTZ zentroan sortutako Itzulbide itzultzaile neuronalari esker, Osakidetzako osasun-langileek txostenak euskaraz idazteko aukera dute azkenean!
Reposted by Eneko Agirre
¡Estamos muy contentas y contentos! Gracias al traductor neuronal creado en HiTZ, las y los sanitarios de Osakidetza podrán escribir sus informes en Euskera!
Reposted by Eneko Agirre
Ai2 @ai2.bsky.social · Aug 28
🎙️ Say hello to OLMoASR—our fully open, from-scratch speech-to-text (STT) model. Trained on a curated audio-text set, it boosts zero-shot ASR and now powers STT in the Ai2 Playground. 👇
Not so bad after all?!
This all confirms very low marginal environmental impacts at the individual level (equivalent to watching streaming video at something like 9-10 seconds of Netflix per prompt at even the higher Mistral numbers), but, of course, this can add up at the societal level for large aggregate impacts.
Reposted by Eneko Agirre
We now have some data on the environmental impact per AI prompt:

Gemini: 0.00024 kWh & 0.26 mL water

ChatGPT: 0.0003 kWh & 0.38 mL

...the same energy as one Google search in 2008 & 6 drops of water. Seems to be improving, too: Google reports a 33x drop in energy use per prompt in a year.

But...
Reposted by Eneko Agirre
The issue with GPT-5 in a nutshell is that unless you pay for model switching & know to use GPT-5 Thinking or Pro, when you ask “GPT-5” you sometimes get the best available AI & sometimes get one of the worst AIs available and you can’t tell and it might even switch within a single conversation.
Reposted by Eneko Agirre
Reposted by Eneko Agirre
We just presented "Truth Knows No Language: Evaluating Truthfulness Beyond English" at ACL 2025! We had some great talks. Thank you! @blanca-calvo-figue.bsky.social @juletxara.bsky.social and @ragerri.bsky.social