Irina Bigoulaeva

I’ll be attending EMNLP 🏖️ in person to present Code Prompting 🧑‍💻 ! I’m super excited about presenting this work! Instead of using plain text, let’s ask LLMs to solve questions using code (without using a code interpreter). x.com/UKPLab/status/…

Ever wondered which input and instructions to use in creative NLG tasks? Even more challenging: how to evaluate them? Check out our #ACL2024NLP paper, where we investigate this for citation text generation! 🧵 (1/7)! Poster: Monday, 11:00 📰: arxiv.org/abs/2407.04046

Very excited to present our paper on *evaluating values and opinions in LLMs* with @vjhofmann at #ACL2024 on Monday / today (oral, 15:15) and Tuesday (poster, 16:00-17:30)! Come by and say hi :)

Are Emergent Abilities in Large Language Models just In-Context Learning? Spoiler: YES 🤯 Through a series of over 1,000 experiments, we provide compelling evidence: arxiv.org/abs/2309.01809 Our results allay safety concerns regarding latent hazardous abilities. A🧵👇 #NLProc

Meet our fellow researchers representing @UKPLab at this year's @ACLmeeting: @IGurevych, @QianRuan_, @J_st_s, @androneil54, @trumancfy, Sheng Lu, @haishuofang, @Furkansahinuc, @KexinWang2049, @LHaausing, @AndreasWaldis, and a very special guest from @BathNLP, @Harish! #ACL2024NLP

Dr Harish Tayyar Madabushi @harish from the Department of Computer Science describes the pros, cons and limitations of large language models (LLMs), such as ChatGPT. 👇 @UniofBathSci #AI #ChatGPT #LLMs

@UKPLab @IBigoulaeva @rachneet4 @IGurevych @BathNLP @CS_TUDarmstadt @TUDarmstadt @UniofBath @ELLISforEurope @Hessian_AI Thanks everyone 🙏for the great conversations and feedback!! This work has been accepted to #ACL2024. If you are in Thailand and would like to chat about Emergent Abilities or LLMs in general, let us know!

🤔 Variational learning is often thought to be impractical 🔥 Plot twist: it actually works better than Adam! Meet IVON, a new optimizer that brings the best out of variational learning – 🧵 (1/9) #NLProc #ICML2024 📰 arxiv.org/abs/2402.17641 youtu.be/TRNYnRRJBRg

🧠💡New preprint! “Linguistic inputs must be syntactically parsable to fully engage the language network” w/ @hee_so_kim, @elizj_lee, NiharikaJhingan @she_jingyuan, @MayaTaliaferro, @LanguageMIT and @ev_fedorenko. Preprint: 📑 tinyurl.com/2hdh6fkh Tweeprint: 🧵👇

Smaller language models (SLMs) are sensitive to input perturbations and fail to generalize. 🚀 Augmenting SLMs with LLM-generated counterfactuals unlocks their potential in extractive QA 🔓 More about our #EACL2024 paper in this 🧵 (1/8) #NLProc 📰 arxiv.org/abs/2309.07822