Founder & CEO of Mosaic AI, NLP PhD, Traveler.
Tel Aviv
Joined August 2009
- Tweets233
- Following121
- Followers658
- Likes556
Pinned Tweet
Can we instantly correct pre-trained models by showing them natural language rules and facts? Can they systematically reason over implicit knowledge while doing so? arxiv.org/abs/2006.06609 New joint work with Oyvind Tafjord, Peter Clark, @yoavgo and @JonathanBerant suggests yes!
Excited to be speaking at #websummit2022 #breakout sessions 1st of Nov. 17:30PM and, 3rd of Nov. 9:45AM, as well as the panel "The socio-economic impacts of AI" at DeepTech stage, at 14:55 Nov 3rd. Come say hello!
Fashionably late, we are happy to announce “CommonsenseQA 2.0: Exposing the Limits of AI through Gamification” arxiv.org/abs/2201.05320 joint work with @oriyoran @Ronan_LeBras @chandra_bhagav @yoavgo @YejinChoinka @JonathanBerant try it out!
Great work by @OriYoran! Generating reasoning examples for RC is made scalable and easy by utilizing Wikipedia tables.
How can we make pre-trained LMs do better reasoning?
In a new work with @AlonTalmor and @JonathanBerant we show that we can automatically generate examples from semi-structured tables and drastically improve performance on RC tasks that involve reasoning.
arxiv.org/abs/2107.07261
Happening in ~12 hours!
When was the painting with two touching fingers completed? We present MultiModalQA: Complex Question Answering over Text, Tables & Images! tinyurl.com/yrkmxnj7 @iclr_conf @OriYoran,@amnoncatav,@LahavDan,@yizhongwyz,@AkariAsai,@gabriel_ilharco, @HannaHajishirzi,@JonathanBerant
We will be presenting @iclr_conf, poster Session 2, May 3, 2021, 9 a.m PTD, come to say hello!
Join @yoavgo, Oyvind Tafjord, Peter Clark @JonathanBerant, and myself in our @NeurIPSConf spotlight QnA in 11 hours - Wednesday 16:10 GMT (tinyurl.com/yxgubcul) and Poster session 17:00 - 19:00 GMT Town A0 - Spot D1 (tinyurl.com/yyfc58wf), would be fun to chat!
Can we instantly correct pre-trained models by showing them natural language rules and facts? Can they systematically reason over implicit knowledge while doing so? arxiv.org/abs/2006.06609 New joint work with Oyvind Tafjord, Peter Clark, @yoavgo and @JonathanBerant suggests yes!
Join us in our QnA session, tomorrow (Wednesday) 9am CET, midnight PDT. Joint work with @yanaiela @yoavgo @JonathanBerant. virtual.2020.emnlp.org/paper… Will be fun!
We present our new year special: “oLMpics - On what Language Model pre-training captures״, arxiv.org/abs/1912.13283, Exploring what
symbolic reasoning skills are learned from an LM objective. We introduce 8 oLMpic games and controls for disentangling pre-training from fine-tuning.
We are excited to announce that our paper “oLMpics - On what Language Model pre-training captures” was accepted to TACL 2020, and will be presented at @emnlp2020! (Question Answering session 11C)
We present our new year special: “oLMpics - On what Language Model pre-training captures״, arxiv.org/abs/1912.13283, Exploring what
symbolic reasoning skills are learned from an LM objective. We introduce 8 oLMpic games and controls for disentangling pre-training from fine-tuning.
We are happy to announce that “Teaching Pre-Trained Models to Systematically Reason Over Implicit Knowledge” will be presented as spotlight at @NeurIPSConf!
Can we instantly correct pre-trained models by showing them natural language rules and facts? Can they systematically reason over implicit knowledge while doing so? arxiv.org/abs/2006.06609 New joint work with Oyvind Tafjord, Peter Clark, @yoavgo and @JonathanBerant suggests yes!
UnifiedQA by @allen_ai, a model based on T5, is the new SOTA on the CommonsenseQA leaderboard (tau-nlp.org/csqa-leaderboard), with an improvement of more than 3 percentage *points* over previous SOTA. Congrats!
We present our new year special: “oLMpics - On what Language Model pre-training captures״, arxiv.org/abs/1912.13283, Exploring what
symbolic reasoning skills are learned from an LM objective. We introduce 8 oLMpic games and controls for disentangling pre-training from fine-tuning.
@facebookai's RoBERTa is up on the CommonsenseQA leaderboard, with 72.1% accuracy, a large improvement over previous results. Exciting! But still ways to go until human performance which is 88.9% tau-nlp.org/csqa-leaderboard
I will be giving a talk on our work “MultiQA: An Empirical Investigation of Generalization and Transfer in Reading Comprehension” tomorrow at 14:10 at 7B Question Answering 3. Come check it out! #ACL2019 #nlproc @ACL2019_Italy
We are happy to announce our new ACL19 paper "MultiQA: An Empirical Investigation of Generalization and Transfer in Reading Comprehension" arxiv.org/abs/1905.13453 a joint work with @JonathanBerant. The repository with a unified format set of datasets used + models coming soon!
Please help us spread the word and make a submission to our shared task and workshop!
Our Call for Papers is out: aclweb.org/portal/content/se…
Regular paper submissions deadline: **August 19th** and shared task submissions deadline: **July 29th**. Please help us spread the word and submit it yourself!
Demo is available at crowdsense.apps.allenai.org , only common sense please not “which company is better” ;)
An app I’ve built to better understand and improve common sense question answering . Try it out! crowdsense.apps.allenai.org/…
Help us develop better common sense #AI! 💡 Add your own creative commonsense questions that you think a child under the age of 7, anywhere around the world, would be able to answer in our new interactive demo CrowdSense: crowdsense.apps.allenai.org/ #CommonSenseAI #AI4Good #NLP
We are happy to announce our new ACL19 paper "MultiQA: An Empirical Investigation of Generalization and Transfer in Reading Comprehension" arxiv.org/abs/1905.13453 a joint work with @JonathanBerant. The repository with a unified format set of datasets used + models coming soon!
Try out our new 6 in-domain and 6 out-of-domain development set datasets for @MRQA2019. Can your model generalize well?
Development datasets released! 6 in-domain and 6 out-of-domain including BioASQ, DROP, DuoRC, RACE, RelationExtraction, TextbookQA! Also released BERT baseline results. All the information at github.com/mrqa/MRQA-Shared-…. Check out and let us know if you have questions! #mrqa2019
