
Amir David Nissan cohen
@amirdnc
ID: 1481572465692315654
13-01-2022 10:24:57
11 Tweet
58 Followers
37 Following

In this work, Amir (and Shauli Ravfogel, and another author who cannot reveal themselves before paper acceptance) shows that diversity is much more important than quantity for model training. u.cs.biu.ac.il/~yogo/diverse-โฆ


Meet ๐๏ธQAmden๐๏ธ your new effective general multi-document model๐๐๐ In our #ACL2023 paper, we propose a new and effective QA-based multi-document pre-training method, specializing in multi-document inputs! w/ Matthew Peters Jacob Goldberger Ido Dagan Arman Cohan 1/n


๐ข New work ๐ With Eran Hirsch @ACL2025 ๐ฆ๐น Daniel Glickman Shauli Ravfogel (((ู()(ู() 'yoav))))๐พ Gal Chechik ๐Demo:huggingface.co/spaces/Royir/Sโฆ ๐Preprint:arxiv.org/abs/2306.08877 We tackle improper binding in diffusion models, where the linguistic structure in the prompt is not reflected in the generation.


We are excited to introduce our EMNLP 2023 findings paper - โNERetrieve: Dataset for Next Generation Named Entity Recognition and Retrievalโ arxiv.org/abs/2310.14282 github.com/katzurik/NERetโฆ Joint w/ Matan Vetzler ,Amir David Nissan cohen and (((ู()(ู() 'yoav))))๐พ


ืืื ืื ืืืื ืืืฉ ืืขืืจืืช Shaltiel ืืื ื ืืจืฆื ื ืืื ืืืืงืืช ืขื ืืืื ืกืืื ืงืืืืื ืืขืืจืืช. ืืืชืืฆืืืช... ืงืฆืช ืืืืจืืช. ืืืืืงืืช ืื: ืืื ืืืจื - ืืืื ืกื ืฉื common sense HeQ - ืืืื ืกื ืฉื ืฉืืืืช ืืชืฉืืืืช ืกื ืืืื ื ืื ืื ื ืจืืืื ืืจืืื ืืฉืืขืืชืืช ืืชืืฆืืืช ืืืืก ืืืืื ืืืงืืจื ืฉื ืืืื. ืืืืื ืขืืื ืืื?


ืชืืื ืืื! ืื ืืื ืื ื ืAmir David Nissan cohen ืืจืฆื ื ืืืืงืืช ืขื ืืืืื ืขื ืืืืกืืื ืงืืืืื ืืขืืจืืช, ืืขืืืื ืืชืืฆืืืช ืงืฆืช ืืืืจืืช: ืคืจืืื ืืื ืืื: V1 ืืืจืฅ ืขื: 1. ืืืจืกืช ืืจื ืกืคืืจืืจื 4.38.1 2. ืืืืง ืืื (ืืื ืงืืื ืืืืฆืื) 3. ืืื BOS V2 ืืืจืฅ ืขื: 1. ืืืจืกืช ืืจื ืกืคืืจืืจื 4.38.2 2. ืืืืง ืืื (ืืื ืงืืื ืืืืฆืื) 3. ืขื BOS


Led by Aviya Maimon, our new paper redefines how we evaluate LLMs. Instead of one flat leaderboard score, we uncover the latent skillsโreasoning, comprehension, ethics, precision & moreโthat really shape LLM ability. Think: psychometrics meets AI. link: arxiv.org/pdf/2507.20208