#MemoryMonday #NLProc "Fake opinion detection: how similar are crowdsourced datasets to real data?" by Fornaciari et al. compares deceptive online reviews collected via crowdsourcing with real fake reviews and find significant differences. milanlproc.github.io/publication/20…

#ThrowbackThursday #NLProc "Beyond Black & White" by Fornaciari et al. proposes a method to leverage annotator disagreement via soft-label multi-task learning for improved performance on various NLP tasks. aclanthology.org/2021.naacl-mai…

#MemoryMonday #NLProc "What the [MASK]? Making Sense of Language-Specific BERT Models" by Nozza et al. presents an overview of language-specific BERT models and their potential compared to multilingual BERT. milanlproc.github.io/publication/20…

#ThrowbackThursday #NLProc "XSTest: A Test Suite for Identifying Exaggerated Safety Behaviours in Large Language Models" by Röttger et al. introduces a new test suite to identify exaggerated safety behaviors in LLMs. arxiv.org/abs/2308.01263

For this week's MilaNLP reading group, Emanuele Moscato presented "Transfusion: Predict the Next Token and Diffuse Images with One Multi-Modal Model" by Chunting Zhou et al. Paper: arxiv.org/pdf/2408.11039 #NLProc

🚨JOB ALERT! If you are working in #NLProc/compSocSci, consider this postdoc call by my colleague Kai Zhu: jobmarket.unibocconi.eu/?id=664 Cool project in a great city (and a chance to work with the folks at MilaNLP). Deadline is Sep 22nd!

For this week's MilaNLP reading group, Himanshu Beniwal presented "Cross-lingual Editing in Multilingual Language Models" by Beniwal et al. Paper: aclanthology.org/2024.findings-… #NLProc

#ThrowbackThursday #NLProc "Conversations as a Source for Teaching Scientific Concepts" by Rooein and Hovy presents a novel and freely available resource for training and evaluating conversation models in educational software using video transcripts. arxiv.org/pdf/2404.10475…

📖For our weekly MilaNLP lab seminar, it was a pleasure to have Rose presenting "Scaling Expertise via Language Models with Applications to Education." #NLProc

#MemoryMonday #NLProc "Measuring Harmful Sentence Completion in Language Models for LGBTQIA+ Individuals" by Nozza et al. investigates the negative impact of LLMs on LGBTQIA+ individuals, finding identity attacks in 13% of completions. aclanthology.org/2022.ltedi-1.4…

✨Many thanks to Università Bocconi for featuring my recent award as Best Young Spanish Researcher in Computer Science, granted by Sociedad Científica Informática de España and Fundación BBVA. Grazie mille! unibocconi.it/it/news/un-pre…

MilaNLP Thank you so much for having me!!! Had a super fun visit and continued visit to the Dolomites 😍

For this week's MilaNLP reading group, Donya Rooein presented "Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers" by CLS et al. Paper: arxiv.org/pdf/2409.04109 #NLProc

#ThrowbackThursday #NLProc "ferret: Framework for Benchmarking Explainers on Transformers" by Attanasio et al. simplifies XAI methods for transformer-based models, visualizes and compares explanations, and evaluates metrics. aclanthology.org/2023.eacl-demo…

Folks in 🇩🇰 & beyond! On Oct 1st, we are hosting (Interacting Minds) a workshop exploring research at the intersection of NLP x social sciences, cognition, & humanities, with 3 awesome keynotes by Dirk Hovy, Maria Antoniak & Sandro Pezzelle. Register here: interactingminds.au.dk/events/single-…

(🧵3/3) 🌀 "Twists, Humps, and Pebbles: Multilingual Speech Recognition Models Exhibit Gender Performance Gaps" by Giuseppe Attanasio, BeatriceSavoldi, Dennis Fucci, and Dirk Hovy, a collaboration with MilaNLP 👉 Preprint: arxiv.org/abs/2402.17954

Two more days to apply!

How does one's religion affect her emotional responses? Many religions cultivate specific emotions, but do LLMs capture this phenomenon? Our #EMNLP2024 Findings paper explores how sacred emotions are represented in current LLMs!

🎉 Our paper "Divine LLaMAs" has been accepted at #EMNLP2024 (Findings)! TL;DR: We uncover LLM biases in emotion attribution across religions, deepening our understanding of the links between religion, culture, and emotions. See you in Miami! 🌴☀️ ⬇️

#MemoryMonday #NLProc "BERTective: Language Models and Contextual Information for Deception Detection" by Fornaciari et al. presents a method for identifying deception that incorporates linguistic context and utilizes BERT's semantic knowledge. aclweb.org/anthology/2021…