🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Synthetic calibration data (for pruning and quantization) generated by the LLM itself is a better approx of the pre-training data dist than "external" data. Really cool work by Miles (Miles Williams) and George (George Chrysostomou) to be presented at #NAACL2025 Link to the paper 👇

I just listened to a thought-provoking Glasgow IR Group #IRTalk from Atsuki Yamaguchi Atsuki Yamaguchi from the University of Sheffield The University of Sheffield on "Cross-lingual Vocabulary Adaptation: Overview and Challenges."

Headed to #ICLR25 , #NAACL, or WWW? Don’t miss the papers from SheffieldNLP, and be sure to say hi to the authors!👋😊

Gutted to miss #NAACL2025 😭 but Miles Miles Williams will be there presenting the following papers: 📄 Main: Self-Calibration for Language Model Quantization and Pruning 📄 RepL4NLP: Vocabulary-Level Memory Efficiency for LM Fine-Tuning Check them out!

Just wrapped up #ICLR2025 and heading to #NAACL2025! I’ll be presenting two papers on detecting model editing: — Oral: Ballroom C, 12:00, April 30 — Poster: May 1, #1618 With so much attention on AI safety at ICLR, these are right on theme. Come say hi! ✨📍🧠 #AISafety #LLMs

More to share soon, but here are the preprints first👇: 🧩 Rulebreakers Challenge arxiv.org/pdf/2410.16502 🛑 Editing Poses Safety Risks arxiv.org/pdf/2502.02958

We're delighted to host Steffen Eger for an insightful talk on "Multimodal Scientific Content Generation with LLMs", covering tools like AutomaTikz, DeTikZify, and ScImage that aim to automate and enhance scientific figure generation🤖📚

Pre-print of SIGIR 2025 #SIGIR2025 paper w/Reem Bin-Hezam “A Generalised and Adaptable Reinforcement Learning Stopping Method” arxiv.org/abs/2505.01907 GRLStop enhances RL-based stopping methods to reduce the effort required to screen collections for relevance.

Join us on 17 & 24 June 2-5 pm CET for 2 workshops online in which we showcase veraAI results. Target audience: #verification, #factchecking and #disinformation detection community, i.e. #journalists, #factcheckers & R&D people. Registration required👇 veraai.eu/posts/two-vera…

Another proud moment: Shun (SWang) has successfully defended his thesis “Interpretable Computational Metaphor Processing”! He produced an impressive body of work in *ACL during his PhD. Big thanks to co-supervisor Po Yang, and to panel members Rob Gaizauskas and Scott Piao🍻🥳

Excited to be talking this evening about the good, the bad, and the ugly of social media at the University's public engagement event "Pint of Science". It's not often I get to give talks in pubs about my research. pintofscience.co.uk/event/say-what…

Excited to share the list of SheffieldNLP accepted papers at ACL 2025 , ICML Conference , and SIGIR 2025 Congrats to all authors, especially our first-time PhD presenters! 🎉 Thread coming soon with details on each paper!

Great day out in the Peaks for the Sheffield NLP Research Away Day just outside Sheffield! 🚍☕🌲 Inspiring talks, thoughtful discussions, and plenty of time to connect. Thanks to all who joined and made it special!🌺

It was such a nice coincidence having our Sheffield NLP Away Day on my name day (21/5)! Glad to be doing research in one of the largest NLP labs in the UK, if not Europe. Also grateful to have the Peak District right at Sheffield's doorstep, especially on such a sunny May day!

Congratulations to Edward Gow-Smith on passing his #PhD #viva with minor corrections! Huge thanks to his supervisor Aline Villavicencio and his viva examiners Professor Dr Hinrich Schütze (Universität München) and Nafise Sadat Moosavi (University of Sheffield)!

1 week left to secure your place at our annual #conference! Join us on 23 June in sunny #Sheffield to hear about our #PhD students’ #AI and #ML #speech and #NLProc research! Cool keynote speakers too! Check out our website for more info! slt-cdt.sheffield.ac.uk/annual-confere…

Super excited to have a cool post-doc lined up for after my PhD, working on Uncertainty Quantification with Nafise Sadat Moosavi at Sheffield NLP, and Benjamin Heinzerling and Kentaro Inui / 乾健太郎 from RIKEN, Japan. Could not be happier 🙂

Can LLMs detect when a question is built on a false premise, especially when it’s hidden in a reasoning step? Not really! We introduce MultiHoax, a benchmark of multi-hop false-premise questions By the brilliant MohammadAmin & HamidReza arxiv.org/abs/2506.00264 #ACL2025 Findings

New paper "Predicting retracted research: a dataset and machine learning approaches" rdcu.be/eqrgY with Aaron Fletcher TLDR: ML can support identification of potentially problematic research