Forough Arabshahi

Research Scientist @Facebook

[Forough Arabshahi @ForoughArabsha1]

(4/4) it learns about it conversationally similar to how we learn from our fellow humans!

We have two poster sessions on Feb 7 from 8:45-10:30 am PST and 4:45-6:30pm PST. Hope to see you there!

[Forough Arabshahi @ForoughArabsha1]

(1/4) Excited to present CORGI (Commonsense Reasoning by Instruction) at #AAAI2021.

We propose a new commonsense reasoning task for conversational agents and a conversational neuro-symbolic reasoning system (CORGI) for solving the task.

https://t.co/XsPSuBmfkD
@tommmitchell

[Forough Arabshahi @ForoughArabsha1]

(3/4) CORGI reasons about human natural language commands: "if-(state holds), then-(perform action), because-(I want to achieve goal)" by outputting a commonsense reasoning chain using its neuro-symbolic logic theorem prover. Whenever CORGI encounters and out of domain example,

[Forough Arabshahi @ForoughArabsha1]

Interested in learning about the recent trends and late-breaking research in human and machine in the loop learning? Join us, tomorrow Dec 12, from 8:15-20:00 PST!

[Divyansh Kaushik @dkaushik96]

Hope to see you all tomorrow @ #NeurIPS2020 HAMLETS workshop starting 8:15AM PT with talks from @tommmitchell @jennwvaughan, @dsweld, @scottyih, @EmmaBrunskill, @ajratner, and others I couldn't find on Twitter.

Also check out our poster sessions at: https://t.co/G7G7spOMd8

[Forough Arabshahi @ForoughArabsha1]

Excited to be a mentor at #WiML2020 talking about Interpretability and explainability in machine learning along with @FinaleDoshi! Join our roundtable Wednesday Dec 9, from 12:30-2:00 pm PST.

[WiML @WiMLworkshop]

We would like to express our sincere gratitude to the inspiring cohort of #WiML2020 MENTORS! We have over 120 mentors from academia & industry! Mentorship roundtables are divided into 3 areas Research (Tables 1–28), Career & Life Advice (Tables 29–50) and Sponsor (Tables 51–63).

[Divyansh Kaushik @dkaushik96]

Call for papers for the #neurips2020 HAMLETS workshop is live along with a great list of invited speakers including @jennwvaughan, @ajratner, @EmmaBrunskill, Sanjoy Dasgupta, Dan Weld, Kristen Grauman, and Finale Doshi-Velez. More to come.

Submit!!!!
https://t.co/g2DRINOlho

[Forough Arabshahi @ForoughArabsha1]

@AnimaAnandkumar @GaryMarcus They show (table 6) that their model doesn't generalize across generators (e.g., 93.6% accuracy drops to 10.9%). In appendix E they discuss this is because the model learns to rely on data artifacts (e.g. sequence length)--result of doing symbolic math without symbol manipulation

[Forough Arabshahi @ForoughArabsha1]

@wzuidema @AnimaAnandkumar @GaryMarcus Interesting work, we got similar results comparing RNN and tree-RNNs/tree-LSTMs here: https://t.co/sFlnrx9IpD but proposed to use different nets for different functions

an updated version of the tree-LSTM extension will be released here tonight: https://t.co/RealLCZxmP

[Hanie Sedghi @HanieSedghi]

Another #Iranian wins the #FieldsMedal ! What an honor for Iran, Kurds, and above that, countries who accept #refugees, and let them fulfill their potential. Humanity is above all wars. #FieldsMedal2018 #proudiranian
https://t.co/25xp7Z1Ycr