Exciting job opportunity! 🧑‍🔬 We at the School of Informatics, University of Edinburgh (School of Informatics, The University of Edinburgh, EdinburghNLP) are looking for a Research Associate (~postdoc) in NLP and ML for a 3-year project fully funded by Huawei via the Huawei-Edinburgh Joint Lab initiative! The project

Our paper on distilling robustness into NLI models will be published at #ACL2024. Joe Stacey is in Bangkok and will be happy to discuss this work. Here is a short video that summarises the paper. #NLProc arxiv.org/abs/2305.13067 youtube.com/watch?v=HWbvBT…

Our #BioNLP2024 paper on predicting the future for patients based on clinical notes will be presented on Friday as part of #ACL2024 workshops. We are forecasting already 2 days after admission what happens to a patient during their whole hospital stay. arxiv.org/abs/2405.11622

Very glad to have presented our work on document-level membership inference (MIAs) for LLMs at USENIX Security ‘24 in Philly. It’s the perfect time to reflect 🪩on our initial work, and on the inherent challenges of MIAs against LLMs (arxiv.org/pdf/2406.17975). 🧵

My amazing collaborators will be presenting 3/3 papers at EMNLP 2024!🚀🚀🚀 EMNLP 2025 Crazy simple KV Cache Compression: arxiv.org/abs/2406.11430 Alessio Devoto Yu Zhao Simone Scardapane Explainable Neuro-Symbolic NLI: arxiv.org/abs/2305.13214 Joe Stacey Oana-Maria Camburu Marek Rei

Here is a little 🧵 about our new #EMNLP2024 paper! arxiv.org/pdf/2305.13214 In this paper we use LLM-generated facts as atoms in atomic inference systems (for NLI). So first of all, what is atomic inference, and why should you care about it?

Here's a special version of our #EMNLP2024 poster for Twitter! 🎉🎉It's an easy read check it out! We introduce an interpretable NLI model with a faithfulness guarantee about which atoms are responsible for each prediction. These atoms are LLM-generated facts from the premise.

68 short summaries of machine learning and NLP research papers. #NLProc #MachineLearning #DeepLearning marekrei.com/blog/68-summar…

A great blog post from Joe Stacey about his research on neural architectures that are inherently interpretable. marekrei.com/blog/creating-…

A great new paper from Lisa Alazraki, factoring LLM tool use into two steps: first generating helpful tool candidates, then reasoning over which tool would be best to use. arxiv.org/abs/2411.04535

Extending pretrained LM-inspired architectures for genome modeling and releasing a tool for predicting epigenetic signals while being cell type-agnostic. Happy to be a co-author on this excellent paper by Alan Murphy , now in Nature Communications. nature.com/articles/s4146… x.com/Al_Murphy_/sta…

Quick intro to myself. I am a CS PhD in LLM Agents Imperial College London with Marek Rei This is our latest work on LLM Agents: StateAct: arxiv.org/abs/2410.02810 (outperforming ReAct by ~10%) Bluesky: bsky.app/profile/ai-nik…

Open position for a Lecturer/Reader in Embodied Natural Language Processing at the School of Informatics (School of Informatics, The University of Edinburgh), University of Edinburgh (The University of Edinburgh) -- apply before Jan 31st, 2025! elxw.fa.em3.oraclecloud.com/hcmUI/Candidat…

The School of Informatics School of Informatics, The University of Edinburgh at the University of Edinburgh The University of Edinburgh is hiring a lecturer or reader in embodied natural language processing. Apply by 31 Jan 2025 at edin.ac/4fqgawg

We are hiring 6 lecturers at Imperial College London, to work on AI, ML, graphics, vision, quantum and software engineering. This includes researchers working on LLMs, #NLProc, generative models and text applications. Deadline 6 Jan. Imperial Computing Imperial NLP imperial.ac.uk/jobs/search-jo…

We have a funded PhD position available at Imperial College London , to research LLMs specialised for healthcare. Co-supervised by Aldo Faisal and myself. Applicable only for UK home students and would have to start already in Feb 2025. If interested, please contact me. Imperial NLP

And we were lucky enough to win the SaTML Conference best paper award!!🌟 Igor Shilov Yves-A. de Montjoye Marek Rei Shubham Jain Manuel Faysse

Many congratulations to Matthieu Meeus Marek Rei and coauthors for the best paper award at SaTML Conference 🎉

Can't wait to see everyone at ICLR and NAACL! Check out some of our awesome papers. Come and say hi, we'd love to have a chat :)

We have a new paper up on arXiv! 🥳🪇 The paper tries to improve the robustness of closed-source LLMs fine-tuned on NLI, assuming a realistic training budget of 10k training examples. Here's a 60 second rundown of what we found!