Happy to share a new PEFT method! Congratulations to all the authors, Alan Ansell Ivan Vulić Hannah Anna Korhonen Edoardo Ponti! 🎉

Happy to share our new preprint! 📜 Congratulations to all the co-authors, Evgeniia Razumovskaia Ivan Vulić and Anna Korhonen! 🥳

🎙Talks talks talks! 🎙 As the new term is just around the corner, we’re happy to invite you to the Easter term Seminar series. Find the schedule below and an up-to-date information with abstract and links at talks.cam.ac.uk/show/index/604….

Happy to share a new NAACL paper! 🎉

🔥New paper!📜 Struggle to align LLM evaluators with human judgements?🤔 Introducing PairS🌟: By exploiting transitivity, we push the potential of pairwise preference in efficient ranking evaluations that has better alignment!🧑‍⚖️ 📖arxiv.org/abs/2403.16950 💻github.com/cambridgeltl/p…

Here’s a kind reminder the talk by Kristina Gligorić from Stanford NLP Group will be happening tomorrow! See you at 4 PM GMT!

Introducing Zero-Shot Tokenizer Transfer (ZeTT) ⚡ ZeTT frees language models from their tokenizer, allowing you to use any model with any tokenizer, with little or no extra training. Super excited to (finally!) share the first project of my PhD🧵

"Role-playing" with LLMs is increasingly popular in chatbots and also "simulation" for social sciences. Can LLMs simulate individual perspectives in subjective NLP tasks? Our #ACL2024 paper w. Nigel Collier investigates this question. 🧵👇 arxiv.org/pdf/2402.10811 1/7

We can’t wait to see you!

Introducing NLLB-LLM2Vec! 🚀 We fuse the NLLB encoder & Llama 3 8B trained w/ LLM2Vec to create NLLB-LLM2Vec which supports cross-lingual NLU in 200+ languages🔥 Joint work w/ Philipp Borchert, Ivan Vulić, and Goran Glavaš during my great research stay at @cambridgeltl

Thrilled to share our new paper: "Can LLM be a Personalized Judge?" We investigate the reliability of LLMs in judging user preferences based on personas and propose improvements using verbal uncertainty estimation to enhance accuracy. 🎭👨‍⚖️ 📄 Paper: arxiv.org/abs/2406.11657

Introducing 🪓Segment any Text! 🪓 A new state-of-the-art sentence segmentation tool! Compared to existing tools (and strong LLMs!), our models are far more: 1. efficient ⚡ 2. performant 🔝 3. robust 🚀 4. adaptable 🎯 5. multilingual 🗺

Attention Instruction: Amplifying Attention in the Middle via Prompting Key findings: 1. LLMs lack relative position awareness 2. We can guide LLM to a specific region with position-based indexing Paper: arxiv.org/pdf/2406.17095 Thanks to: Zaiqiao Meng and Nigel Collier