Sample ManyNames images with associated names, in English and Mandarin Chinese
Releasing v. 2.3 of ManyNames, an object naming dataset with 25K objects in real world images (English, plus partial coverage in Catalan and Mandarin Chinese). Check it out!
amore-upf.github.io/manynames/
(New in this version: further data cleaning, speaker ID, more lexical info)
@ecesuurker.bsky.social presenting NeLLCom-Lex: A Neural-agent Framework to Study the Interplay between Lexical Systems and Language Use (Zhang et al, EMNLP Findings, 2025) !
LLMs as a synthesis between symbolic and distributed approaches to language (ACL Findings, 2025), a talk by Gemma Boleda @gboleda.bsky.social
Our group presented our work at Deep Learning BCN! Some highlights below.
@dlbcnai.bsky.social
Amazing work by Jeanne Bruneau-Bongard, Emmanuel Chemla, and Thomas Brochhagen!
Many forces have been argued to shape natural language lexica, and there are different ways they can be operationalized and interact. We study which out of a set of forces and their interactions best fit cross-linguistic data. Now out in Cognitive Science: onlinelibrary.wiley.com/doi/10.1111/...
Great work by Xixian Liao, Thomas Brochhagen, @gboleda.bsky.social and @laiamayol.bsky.social !
Do you use a pronoun more often when the entity you’re talking about is more predictable?
Previous work offers diverging answers so we conducted a meta-analysis, combining data from 20 studies across 8 different languages.
Now out in Language: muse.jhu.edu/article/969615
📢 Seminari de recerca organitzat pel COLT- URLING, "LLM and human language: representations, judgments, and historical change".
📆 29/09/2025
🕦 15:30
🎤 Adele Goldberg (Princeton University)
🚩55.410, Edifici Tànger del Campus Poblenou - UPF
ℹ️ ja.cat/wi2t7
@colt-upf.bsky.social
📢 Seminari de recerca organitzat pel COLT- URLING, "Associative memory in psycholinguistics and in AI architectures".
📆 01/10/2025
🕦 12:00
🎤 Jakub Dotlačil
🚩55.410, Edifici Tànger del Campus Poblenou - UPF
ℹ️ ja.cat/U5xH2
@colt-upf.bsky.social
Sigmoid function. Non-linearities in neural network allow it to behave in distributed and near-symbolic fashions.
New paper! 🚨 I argue that LLMs represent a synthesis between distributed and symbolic approaches to language, because, when exposed to language, they develop highly symbolic representations and processing mechanisms in addition to distributed ones.
arxiv.org/abs/2502.11856
📢I am hiring a Postdoc to work on post-training methods for low-resource languages. Apply by August 15 employment.ku.dk/faculty/?sho....
Let's talk at #ACL2025NLP in Vienna if you want to know more about the position and life in Denmark.
Screenshot of first page of paper. It is here: https://arxiv.org/pdf/2507.00828 Abstract: Topic model and document-clustering evaluations either use automated metrics that align poorly with human preferences or require expert labels that are intractable to scale. We design a scalable human evaluation protocol and a corresponding automated approximation that reflect practitioners' real-world usage of models. Annotators -- or an LLM-based proxy -- review text items assigned to a topic or cluster, infer a category for the group, then apply that category to other documents. Using this protocol, we collect extensive crowdworker annotations of outputs from a diverse set of topic models on two datasets. We then use these annotations to validate automated proxies, finding that the best LLM proxies are statistically indistinguishable from a human annotator and can therefore serve as a reasonable substitute in automated evaluations
Evaluating topic models (and document clustering methods) is hard. In fact, since our paper critiquing standard evaluation practices four years ago, there hasn't been a good replacement metric
That ends today (we hope)! Our new ACL paper introduces an LLM-based evaluation protocol 🧵
🎉New paper "Prediction Hubs are Context-Informed Frequent Tokens in LLMs" from our lab, accepted at ACL 2025!
If you're interested in representational geometry, come find Beatrix Nielsen and Marco Baroni at the poster :)
Today at UPF Campus de la Ciutadella at 2:30 pm! Come slightly earlier to check in!
Sala Polivalent 24S18
maps.app.goo.gl/n1hBxiviKcLW...
📢 𝗟𝗼𝗰𝗮𝘁𝗶𝗼𝗻 𝗰𝗵𝗮𝗻𝗴𝗲📢
UPF Campus de la Ciutadella
**Sala Polivalent 24.S18**
Thank you for bearing with us!
Last day to sign up for the COLT Symposium!
Register: tinyurl.com/colt-register
📢 𝗟𝗼𝗰𝗮𝘁𝗶𝗼𝗻 𝗰𝗵𝗮𝗻𝗴𝗲📢
June 2nd, 14:30 - 19:00
UPF Campus de la Ciutadella
Room 40.101
maps.app.goo.gl/1216LJRsWmTE...
⭐ Registration open til May 27th! ⭐
Website: www.upf.edu/web/colt/sym...
June 2nd, UPF
𝗦𝗽𝗲𝗮𝗸𝗲𝗿 𝗹𝗶𝗻𝗲𝘂𝗽:
Arianna Bisazza (language acquisition with NNs)
Naomi Saphra (emergence in LLM training dynamics)
Jean-Rémi King (TBD)
Louise McNally (pitfalls of contextual/formal accounts of semantics)
Updated website: www.upf.edu/web/colt/sym...
𝗚𝗲𝘁𝘁𝗶𝗻𝗴 𝘁𝗵𝗲𝗿𝗲:
𝗪𝗵𝗲𝗻: 2nd June 2025, 14:30 - 19:00
𝗪𝗵𝗲𝗿𝗲: UPF Poblenou, Auditori (enter via Roc Boronat building) maps.app.goo.gl/2WMt21hR5L9r...
In-person only, with mandatory registration:
tinyurl.com/colt-register
See you there!
🧵3/3
Our speakers span a wide range of expertise between AI, linguistics, and neuroscience.
14:30 Arianna Bisazza (Uni. Groningen)
15:30 Naomi Saphra (Harvard)
-- coffee break --
17:00 Jean-Rémi King (Meta AI)
18:00 Louise McNally (UPF)
Abstracts: tinyurl.com/colt-site
🧵2/3
Announcing the COLT Symposium on June 2nd!
𝗘𝗺𝗲𝗿𝗴𝗲𝗻𝘁 𝗳𝗲𝗮𝘁𝘂𝗿𝗲𝘀 𝗼𝗳 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗶𝗻 𝗺𝗶𝗻𝗱𝘀 𝗮𝗻𝗱 𝗺𝗮𝗰𝗵𝗶𝗻𝗲𝘀
What properties of language are emerging from work in experimental and theoretical linguistics, neuroscience & LLM interpretability?
Info: tinyurl.com/colt-site
Register: tinyurl.com/colt-register
🧵1/3
Please find us at #ICLR2025! We will present our work on intrinsic dimension as a cue for stages of language processing in LLMs.
Saturday morning, Poster session 5
Hall 3 + Hall2B #563
iclr.cc/virtual/2025...
Arxiv: arxiv.org/abs/2405.15471
📢 Upcoming Seminar
Words are weird? On the role of lexical ambiguity in language
🗣 Gemma Boleda (Universitat Pompeu Fabra, Spain)
Why is language so ambiguous? Discover how ambiguity balances cognitive simplicity and communicative complexity through large-scale studies.
📍 UniMiB, Room U6-01C, Milan
⚡New position paper from Gemma Boleda: is it time to make peace between symbolic and continuous approaches to language?
The project I did with Marco Baroni and Iuri Macocco while I was in Barcelona is now on Arxiv: arxiv.org/abs/2502.10201 🎉
TLDR below 👇
This year, CoNLL will be accepting *non-archival* (as well as archival) submissions! www.conll.org #CoNLL2025
Follow CoNLL at
@conll-conf.bsky.social
Here's our work accepted to #ICLR2025!
We look at how intrinsic dimension evolves over LLM layers, spotting a universal high-dimensional phase.
This ID peak is where:
- linguistic features are built
- different LLMs are most similar,
with implications for task transfer
🧵 1/6
Què és l’aprenentatge profund ?
La @marionamec.bsky.social de @neurofregides.bsky.social ens ho explica en motiu del Deep Learning Barcelona Symposium 2024 (@dlbcn.ai), aquest dijous 19 de desembre.
#deeplearning #ciencia #català #barcelona
www.youtube.com/shorts/R4u_Z...
Conclusion: for communication in-context,
Lexical systems with a soft mapping between referents and names let speakers maximize communication accuracy while minimizing complexity.
Paper: aclanthology.org/2024.emnlp-m...
3/3