Lightnews — Scholar-powered news

Reposted by GroNLP

Federico Pianzola @fpianz.bsky.social · 5d

Great tips from @jiruiqi.bsky.social about how to choose your reinforcement learning strategy. @gronlp.bsky.social

Slide showing practical tips about how to choose reinforcement learning strategy between PPO, DPO, and GRPO.

1 6

GroNLP @gronlp.bsky.social · 7d

Interviews with selected candidates will be held on-line between 10th–14th November 2025.

GroNLP @gronlp.bsky.social · 7d

You would be joining a vibrant environment embedded in the 𝐂𝐞𝐧𝐭𝐞𝐫 𝐟𝐨𝐫 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐚𝐧𝐝 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐨𝐧 (𝐂𝐋𝐂𝐆) at the University of Groningen.⁣ You can apply for this position until 𝐎𝐜𝐭𝐨𝐛𝐞𝐫 𝟏𝟐 𝟐𝟎𝟐𝟓 Dutch local time (CET) .⁣

1

GroNLP @gronlp.bsky.social · 7d

𝐓𝐡𝐞 𝐈𝐍𝐂𝐑𝐄𝐂 𝐩𝐫𝐨𝐣𝐞𝐜𝐭 (𝐄𝐑𝐂 𝐂𝐨𝐆, 𝐏𝐈 𝐀𝐧𝐚 𝐆𝐮𝐞𝐫𝐛𝐞𝐫𝐨𝐟) 𝐢𝐬 𝐥𝐨𝐨𝐤𝐢𝐧𝐠 𝐟𝐨𝐫 𝐚 𝐧𝐞𝐰 𝐭𝐞𝐚𝐦 𝐦𝐞𝐦𝐛𝐞𝐫! ⁣🎉
⁣
We have an opening for a 𝐏𝐨𝐬𝐭𝐝𝐨𝐜 𝐩𝐨𝐬𝐢𝐭𝐢𝐨𝐧 in 𝐂𝐫𝐞𝐚𝐭𝐢𝐯𝐢𝐭𝐲, 𝐓𝐫𝐚𝐧𝐬𝐥𝐚𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐓𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐲 for a starting duration of 12 months extendible to 30 months.⁣

For further details -> www.rug.nl/about-ug/wor...

Vacatures bij de RUG

www.rug.nl

1 5 4

GroNLP @gronlp.bsky.social · 12d

2

GroNLP @gronlp.bsky.social · 12d

𝗧𝗵𝗲 𝗘𝘂𝗿𝗼𝗽𝗲𝗮𝗻 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗲𝗿𝘀’ 𝗡𝗶𝗴𝗵𝘁 𝘄𝗮𝘀 𝗮𝗻𝘁𝗶𝗰𝗶𝗽𝗮𝘁𝗲𝗱 𝗯𝘆 𝗮𝗻 𝗲𝗾𝘂𝗮𝗹𝗹𝘆 𝗲𝘅𝗰𝗶𝘁𝗶𝗻𝗴 𝗮𝗳𝘁𝗲𝗿𝗻𝗼𝗼𝗻! 😍
Here are some photos of our GroNLP station at the Forum, where we welcomed high school classes and potential future researchers. 🧑‍🔬🤖
Quizzes, puzzles, demos, card games, and of course, lots of chocolate! 🍫

1 7 14

GroNLP @gronlp.bsky.social · 26d

1 5

GroNLP @gronlp.bsky.social · 26d

Today we were at #CLIN35 hosted by @kuleuvenuniversity.bsky.social - as always: great interactions, meeting old friends/colleagues, and lots of presentations from us @facultyofartsug.bsky.social

2 8 22

GroNLP @gronlp.bsky.social · 28d

We recharged our batteries with a wonderful weekend together in the province of Drenthe! 🌳🍃
As per tradition, we kick off the academic year with plenty of sharing, smiles, and sports 💘

We wish everyone a wonderful start of the academic year <3 may it be inspiring and exciting 🎓

5 16

Reposted by GroNLP

Ana Guerberof @anaguerberof.bsky.social · 29d

#INCREC is looking for a new team member:

Postdoc in #Creativity, #Translation and #Technology.
Come an join the team at University of Groningen!

Applications until 12 October 11:59pm, 2025 Dutch local time (CET) .

All information 👇

www.rug.nl/about-ug/wor...

Vacatures bij de RUG

www.rug.nl

3 3

GroNLP @gronlp.bsky.social · 28d

In this episode, we had a conversation with 𝗠𝗲𝗿𝗶𝘁𝘅𝗲𝗹𝗹 𝗦𝗮𝗹𝘃𝗮𝗻𝘆 𝗕𝗮𝗹𝗮𝗱𝗮, an experienced literary translator from Scandinavian languages to Catalan, about the translator’s role in-between the author and the reader, meta-creativity and the impact new technologies might have on translation.
Check it out!

1

GroNLP @gronlp.bsky.social · 28d

𝗔 𝗻𝗲𝘄 𝗲𝗽𝗶𝘀𝗼𝗱𝗲 𝗼𝗳 𝗼𝘂𝗿 𝗜𝗡𝗖𝗥𝗘𝗖 𝘁𝗲𝗮𝗺'𝘀 𝗽𝗼𝗱𝗰𝗮𝘀𝘁 “𝗧𝗵𝗲 𝗩𝗶𝘀𝗶𝗯𝗹𝗲 𝗔𝗿𝘁 𝗼𝗳 𝗧𝗿𝗮𝗻𝘀𝗹𝗮𝘁𝗶𝗼𝗻” 𝗶𝘀 𝗼𝘂𝘁! 😍

#creativity #xl8n #MT #literature #INCREC @gronlp.bsky.social @facultyofartsug.bsky.social

S2E08: Meritxell Salvany Balada "A translators main work is not so much to create something but to reproduce it in a way that doesn't crumble"

open.spotify.com

1 3

GroNLP @gronlp.bsky.social · Aug 4

#ACL2025 has just finished and it was great to meet with old friends and connect with new colleagus. In case you missed our papers/presentations, this was us 🐮👇
@facultyofartsug.bsky.social

4 10

GroNLP @gronlp.bsky.social · Jun 20

We’re excited to share a new addition from our lab to the family of language-specific BLiMPs, this time the morphologically rich Turkish language is in the spotlight! 🎊✨

Arianna Bisazza @arianna-bis.bsky.social · Jun 19

Proud to introduce TurBLiMP, the 1st benchmark of minimal pairs for free-order, morphologically rich Turkish language!

Pre-print: arxiv.org/abs/2506.13487

Fruit of an almost year-long project by amazing MS student @ezgibasar.bsky.social in collab w/ @frap98.bsky.social and @jumelet.bsky.social

TurBLiMP: A Turkish Benchmark of Linguistic Minimal Pairs

We introduce TurBLiMP, the first Turkish benchmark of linguistic minimal pairs, designed to evaluate the linguistic abilities of monolingual and multilingual language models (LMs). Covering 16 linguis...

arxiv.org

5

GroNLP @gronlp.bsky.social · May 31

And once again 🎊 a little read for you all to enjoy :)

Check it out !!

Gabriele Sarti @gsarti.com · May 30

📢 New paper: Can unsupervised metrics extracted from MT models detect their translation errors reliably? Do annotators even *agree* on what constitutes an error? 🧐

We compare uncertainty- and interp-based WQE metrics across 12 directions, with some surprising findings!

🧵 1/

4

Reposted by GroNLP

Francesca Padovani @frap98.bsky.social · May 30

“Child-Directed Language Does Not Consistently Boost Syntax Learning in Language Models”

I’m happy to share that the preprint of my first PhD project is now online!

🎊 Paper: arxiv.org/abs/2505.23689

Child-Directed Language Does Not Consistently Boost Syntax Learning in Language Models

Seminal work by Huebner et al. (2021) showed that language models (LMs) trained on English Child-Directed Language (CDL) can reach similar syntactic abilities as LMs trained on much larger amounts of ...

arxiv.org

2 17 62

GroNLP @gronlp.bsky.social · May 23

🎊 New work out from our brilliant PhDs 🎊

Gabriele Sarti @gsarti.com · May 23

Prev. work 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 thread! ⬇️

Paper: arxiv.org/abs/2505.16612

2 8

Reposted by GroNLP

Federico Pianzola @fpianz.bsky.social · May 22

A recent @anthropic.com publication has a section on how LLMs plan rhymes when generating poems.

Thanks @gsarti.com @gronlp.bsky.social for the tip!

transformer-circuits.pub/2025/attribu...

On the Biology of a Large Language Model

We investigate the internal mechanisms used by Claude 3.5 Haiku — Anthropic's lightweight production model — in a variety of contexts, using our circuit tracing methodology.

transformer-circuits.pub

1 4

Reposted by GroNLP

Federico Pianzola @fpianz.bsky.social · May 15

Tomorrow, Friday 16th May, I will present the GOLEM project at KAIST in South Korea. Online too.

2025-05-16 13:30~15:00 KST (15th May, 21:30 Pacific time)
The GOLEM Project: Graphs and Ontologies for Literary Evolution Models

DM me for Zoom link

#DigitalHumanities @gronlp.bsky.social

introductory slide of a talk by Federico Pianzola, titled "Graphs and Ontologies for Literary Evolution Models", a project funded by the European Union

1 2 5

GroNLP @gronlp.bsky.social · May 14

Few days back 🍦 celebrating spring, sunshine, and one of our most cherished lab member! Nothing like ice cream and an indulgent break to mark a birthday the Computational Linguistics way ☀️🎉 #GroNLP #ComputationalLinguistics #SpringVibes

1 8

Reposted by GroNLP

Giulia Rambelli @grambelli.bsky.social · Apr 25

🎓Already 3 weeks at @rug.nl
Today I joined my first @gronlp.bsky.social reading group, where prithamajumdar.bsky.social shared her NAACL paper on comparisons & similes - couldn’t help but think of @abstractionerc.bsky.social when seeing this slide❤️

📢 Next week it’s my turn to present—stay tuned!

1 7

Reposted by GroNLP

MariannaBolog @mariannabolog.bsky.social · May 1

@grambelli.bsky.social these weeks is hosted by @gronlp.bsky.social for a postdoctoral visiting sponsored by @abstractionerc.bsky.social for a collaboration with @tommasoc80.bsky.social and @piasommerauer.bsky.social on the emergence of language-mediated stereotypes, a form of overgeneralization.

4 7

Reposted by GroNLP

Giulia Rambelli @grambelli.bsky.social · May 2

📢 Excited to present my postdoc research within @abstractionerc.bsky.social at the @gronlp.bsky.social reading group (@facultyofartsug.bsky.social) TODAY!
🤖🧠 I'll explore how LLMs interpret generics and compounds —and what this tells us about their #generalization abilities.
#NLP #LLMs #linguistics

5 5

Reposted by GroNLP

Jaap Jumelet @jumelet.bsky.social · Apr 7

✨New paper ✨

Introducing 🌍MultiBLiMP 1.0: A Massively Multilingual Benchmark of Minimal Pairs for Subject-Verb Agreement, covering 101 languages!

We present over 125,000 minimal pairs and evaluate 17 LLMs, finding that support is still lacking for many languages.

🧵⬇️

3 22 75

GroNLP @gronlp.bsky.social · Apr 8

Thrilled to announce the release of MultiBLiMP 1.0 by our team at GroNLP Lab! 🙆‍♀️

Arianna Bisazza @arianna-bis.bsky.social · Apr 8

Modern LLMs "speak" hundreds of languages... but do they really?
Multilinguality claims are often based on downstream tasks like QA & MT, while *formal* linguistic competence remains hard to gauge in lots of languages

Meet MultiBLiMP!
(joint work w/ @jumelet.bsky.social & @weissweiler.bsky.social)

Jaap Jumelet @jumelet.bsky.social · Apr 7

✨New paper ✨

Introducing 🌍MultiBLiMP 1.0: A Massively Multilingual Benchmark of Minimal Pairs for Subject-Verb Agreement, covering 101 languages!

We present over 125,000 minimal pairs and evaluate 17 LLMs, finding that support is still lacking for many languages.

🧵⬇️

2 9