New blog post experimenting with using semantic search to find potential machine learning models for fine-tuning on the Hugging Face model hub. danielvanstrien.xyz/huggingface/hu…

🔢Semantic search based on dense vector spaces can significantly improve search results, but generalizes badly to new domains & new language ☹️ 🔍In this talk, I will give an outlook into the next generation of algorithms we will be using to improve search results.

Join today at 7pm CEST (UTC+2) on a webinar on how to improve search using Neural Networks: - Overview of neural architectures for search - Pros and cons - Distillation across architectures

I'm super happy to share that I have joined Google research this fall as a PhD research intern in California to work on multilingual search w/ Daniel Cer, Jianmo Ni! 😍🥳 It's my first time in the US. If you are in the bay area, let's meet! 🇺🇸🍻 #google #internship #NLP #IR

This latest release for Transformers Interpret (v0.8.1) brings some features I'm really excited about particularly the new pairwise classification explainer. (1/8) github.com/cdpierse/trans…

Activation functions reduce the topological complexity of data. Best AF may be diff for diff models and diff layers, but most Transformer models use GELU. What if the model learns optimized activation functions during training? led by Haishuo with Ji Ung Lee and Iryna Gurevych

Efficient Few-Shot Learning Without Prompts abs: arxiv.org/abs/2209.11055

I am super proud to share the work of my Google AI internship 🎉 FiD-Light is an efficient retrieval-augmented generation model advancing the state-of-the-art effectiveness on six KILT tasks considerably 🙌 w/ Jiecao Chen Karthik Raman Hamed Zamani 📄 arxiv.org/abs/2209.14290

MTEB - Massive Text Embedding Benchmark 🧨 Text embeddings are usefull for many applications 💻, but still their evaluation is often done rather poorly on trivial datasets 🙁. MTEB is here to change it. We collected 58 datasets across 8 tasks and evaluated many public models.

New paper 🚨 arxiv.org/abs/2211.09260 Can we train a single search system that satisfies our diverse information needs? We present 𝕋𝔸ℝ𝕋 🥧 the first multi-task instruction-following retriever trained on 𝔹𝔼ℝℝ𝕀 🫐, a collections of 40 retrieval tasks with instructions! 1/N

#ChatGPT and #GPT3 are hot. But let’s be practical, when we want to reproduce GPT-3 or use it in our applications. Why did all of the public reproduction of GPT-3 fail? In which tasks should we use GPT-3.5/ChatGPT? I tried to answer them in a new blog: jingfengyang.github.io/gpt .

Happy to collaborate with Yongxin Huang . Congrats!

Congrats to @anxo6ila🎉!

🇺🇳𝟐𝟓𝟎𝐌 𝐖𝐢𝐤𝐢𝐩𝐞𝐝𝐢𝐚 𝐢𝐧𝐭𝟖 & 𝐛𝐢𝐧𝐚𝐫𝐲 𝐄𝐦𝐛𝐞𝐝𝐝𝐢𝐧𝐠𝐬 🇺🇳 int8 & binary embeddings are amazing: 💰 Up to 100x lower cost 🚀 Up to 30x faster 🏆 State-of-the-art search quality Wikipedia embedded in 300+ languages: huggingface.co/datasets/Coher…

Andreas Waldis HSLU Hochschule Luzern Yufang Hou IBM Research Iryna Gurevych Md. Imbesat Hassan Rizvi Xiaodan Zhu ECE Queens Ingenuity Labs Indraneil Paul Goran Glavaš Universität Würzburg #UniWürzburg Jonibek Mansurov Jinyan Su (on job market) Artem Shelmanov Akim Tsvigun Osama Afzal Alham Fikri Aji @nyhabash Preslav Nakov MBZUAI TU Darmstadt @IstiCnr NYU Abu Dhabi Furkan Şahinuç Ilia Kuznetsov Irina Bigoulaeva rachneet Harish Tayyar Madabushi BathNLP Justus-Jonas Erker 🇪🇺 @ ACL2024 🇹🇭 Florian Mai Computer Science | KU Leuven Nils Reimers cohere Jerry Spanakis 🟥 🦋 gerasimoss.bsky.social Maastricht University Max Glockner Qian Ruan »DAPR: A Benchmark on Document-Aware Passage Retrieval« by Kexin Wang (UKP Lab), Nils Reimers (cohere) and Iryna Gurevych (12/🧵) #ACL2024NLP 📑 arxiv.org/abs/2305.13915

𝐒𝐞𝐦𝐚𝐧𝐭𝐢𝐜 𝐒𝐞𝐚𝐫𝐜𝐡 𝐨𝐧 𝟏𝟎𝟎𝐌 𝐝𝐨𝐜𝐬 - 𝐖𝐢𝐭𝐡 𝟏𝟎𝟎𝐌𝐁 𝐨𝐟 𝐌𝐞𝐦𝐨𝐫𝐲 GPU-poor and Memory-poor, and not having 500GB of memory to embed & index 100M docs? Still want to participate at TREC-RAG 2024? Introducing 𝐃𝐢𝐬𝐤𝐕𝐞𝐜𝐭𝐨𝐫𝐈𝐧𝐝𝐞𝐱

I am attending acl24 at Bangkok🇹🇭 to present my paper DAPR huggingface.co/datasets/UKPLa… (Mon 14:00 poster). Looking forward to meeting you! #ACL2024NLP

I’m presenting the DAPR paper at ACL’24 poster session 2 14:00-15:30. Welcome to conversations!

Meet our fellow researchers representing UKP Lab at this year's @ACLmeeting: Iryna Gurevych, Qian Ruan, Justus-Jonas Erker 🇪🇺 @ ACL2024 🇹🇭, Indraneil Paul, Fengyu Cai, Sheng Lu, Haishuo, Furkan Şahinuç, Kexin Wang, Haau-Sing Li 李 效丞, Andreas Waldis @EMNLP2024, and a very special guest from BathNLP, @Harish! #ACL2024NLP

1/12 📢New paper alert “DOCE: Finding the Sweet Spot for Execution-Based Code Generation” The work is a group effort with Patrick Fernandes, Iryna Gurevych, and Andre Martins. UKP Lab ELLIS Unit Lisbon DeepSPIN 📰: arxiv.org/pdf/2408.13745