מורן מזרחי על יצירתיות חישובית. בינה מלאכותית ליצירת מתכונים

🚨New #NAACL2024 main 🇲🇽 Paper Alert 🚨 Happy to share our work arxiv.org/abs/2403.01139 🥳 A new pipeline, ParallelPARC ✅ A new benchmark, ProPara-Logy ✅ LLMs & humans evaluation ✅ w. Yonatan Bitton Ron Yosef & our Guru🧠Hyadata Lab (Dafna Shahaf)!

🚨New Dataset Distillation paper 🚨 We propose PoDD, a new dataset distillation setting for a tiny🤏under 1 image-per-class (IPC) budget❗In this CIFAR100 example, current SoTA is 35.5% acc using ~100k pixels, PoDD gets 35.7% with ~40k pixels Project: vision.huji.ac.il/podd 🧵👇

🐣Too Long; Didn't Read: Most LLMs cite full documents for support, but they're too long so people don't actually read them. In our new work "Attribute First, then Generate”, we introduce fine-grained attributions, where LLMs are required to highlight only relevant information

The day when we'll be able to talk with LLMs is ever closer! 🗣️ However, in a real conversation, we expect an immediate response. ⏰ To reduce the response latency, we propose LLM2Speech, which synthesizes speech from an LLM while it is generating text. IBM Research #ICASSP2024

🎉Excited to share our paper When LLMs are Unfit, Use FastFit: Fast and Effective Text Classification with Many Classes was accepted to NAACL HLT 2024! 🚀SOTA results ⚡Fast training & inference 🎯High accuracy 📄Paper:arxiv.org/abs/2404.12365 💻Package:github.com/IBM/fastfit

🚨I am excited to share that our paper “Multi-Review Fusion-in-Context” has been accepted to #NAACL24 findings. 🇲🇽 🔗paper: arxiv.org/abs/2403.15351 🔗project website: fusereviews.github.io work done with Ori Shapira Ran Levy and Ido Dagan during an internship at Amazon 🧵(1/n)

Exciting news! Our paper: "A Nurse is Blue and Elephant is Rugby: Cross Domain Alignment in Large Language Models Reveal Human-like Patterns" was accepted to #CogSci2024! 🎉 We explore how LLMs map concepts across domains, and reason about those mappings. 🧠✨

PromptEval: multi-prompt evaluation made practical! PromptEval efficiently estimates the performance distribution of LLMs over a big set of prompts (>100) using less than 5% of the compute🦾 Paper: arxiv.org/abs/2405.17202 GitHub: github.com/felipemaiapolo… Thread below🧵1/8

Moran Mizrahi Guy Kaplan Dan H.M 🎗 Rotem Dror Hyadata Lab (Dafna Shahaf) Gabriel Stanovsky Thanks for making your data available, Moran and team! We have built from your work :) Now, we can compute performance summary statistics across prompts efficiently x.com/felipemaiapolo…

Do you like yellow? Then, according to LLMs, you are probably a school bus driver! Excited to share our new paper about Semantic Leakage in Language Models! Joint work with my wonderful collaborators @terra Alisa Liu luke Noah A. Smith Paper: gonenhila.github.io/files/Semantic… 1/10

1/5 Moran Mizrahi will present her work on the brittleness of LLM evaluation to prompt paraphrases on Aug 12th at 12:15PM in World Ballroom A (Talk) and on Aug 13th at 4PM in Convention Center A1 (Poster). x.com/moranmiz/statu…

🚀 Lots of exciting work from the HUJI NLP group and our collaborators at #ACL2024! Come talk to us about it! Moran Mizrahi Gili Lior @ ACL 2024 Itay Itzhak Gabriel Stanovsky

The issue of prompt sensitivity comes up often in practice, but its comprehensive studies can be hard to come by. If you wish to learn more, here are several excellent works on this topic: x.com/melaniesclar/s… x.com/moranmiz/statu… huggingface.co/blog/open-llm-…

Attending #ACL2024? 🎤Don't miss my presentation on our TACL paper about the sensitivity of LLMs to prompt paraphrases! 🗣️ Oral: Mon, 12/8, 12:15 PM (World A) 📊 Poster: Tue, 13/8, 16:00 PM Details below ⬇️ w/ Guy Kaplan, Dan H.M 🎗, Rotem Dror, Hyadata Lab (Dafna Shahaf), Gabriel Stanovsky

Excited to have presented our TACL paper about the sensitivity of LLMs to prompt paraphrases at #ACL2024! 📊✨ Guy Kaplan, Dan H.M 🎗, Rotem Dror, Hyadata Lab (Dafna Shahaf), Gabriel Stanovsky direct.mit.edu/tacl/article/d…