People are increasingly worried about backdoors in LLMs…
so we developed a method to remove them.
New paper: Removing the NEEDLE in the Haystack: Backdoor Removal in LLMs via Weight Orthogonalisation
Congratulations to Minoo Kim for leading this work and @lampos
Paper: arxiv.org/abs/2610.00348
Code: github.com/LocaiLabs/NEEDLE
Models: huggingface.co/collections/l…
Really proud to share that the GB1 app is now live on iOS and Android!
We’re also on @ProductHunt today, currently sitting right behind @nvidia's Nemoclaw. Grateful for every upvote and message of support!
Big props to the @locai_labs team!
George Drayson retweeted
The name’s GB1. The AI Assistant with a licence to be private, independent, sustainable, and most importantly, British.
Will you answer the call?
Excited to be presenting our paper “Machine-generated text detection prevents language model collapse” at EMNLP 2025!
🗓️ I’ll be presenting our work on Friday, Nov 7 10:30–12:00 in the Poster Session in Hall C.
#EMNLP2025 #LLMs #NLP #AI
In this work, we analyse how LLM decoding strategies influence collapse dynamics and propose a detector-based resampling method to prevent model collapse.
Paper link: aclanthology.org/2025.emnlp-…
Our paper “Machine-generated text detection prevents language model collapse” has been accepted for presentation at EMNLP 2025 Main Conference
Huge thanks to @lampos and Emine Yilmaz for their supervision. Excited to present in Suzhou and connect with the NLP community!










