Daniel Khashabi 🕊️ @DanielKhashabi Twitter profile

Pinned Tweet

Daniel Khashabi 🕊️

@DanielKhashabi

2 months

Looking to benchmark your web-based AI agent? Consider TurkingBench! See @KevLXu 's post for details.

Kevin Xu

@KevLXu

2 months

Ever wondered how LLMs stack up against human crowdsource workers? I'm thrilled to share "TurkingBench", a benchmark of web-based tasks for multi-modal and interactive AI agents. Draft: Project: Code:

1

14

38

0

6

19

Last Seen Profiles

@googletokki

@pratim_vas_samo

@Megatronstaint

@clubb32

@DarceyEdkins

@ARKEAULTIMCHALL

@AgustinLebron3

@wlals_rnr

@vveb1og

@DuperFPS

@GrandmamaAngela

@saiteuuu

@Floricra

@BruceHillMelb

@sm2_r7

@kristynnana

@stwmaniax

@sliceofseokmin

@clayyytonbigsby

@kissy6476540782

@liza_pflaum

@Midnightcause

@JonnyThelander

@Tango21IX

@saiko_freaky

@lovelyzoe77

@an_liberte

@auddryyy

@iCarryNeutron

@peytonoleary6

@PetraLaiti

@QsmpEN

@gerardfagway

@d1amond_shot

@SPEnergyNetwork

@swiftyyfps

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

Life update: Thrilled to announce that I will join Johns Hopkins University @jhuclsp @jhucompsci @JohnsHopkins as an assistant professor of computer science in the fall! This is an honor of a lifetime and I will do my best to rise to the occasion.

82

24

560

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

For my first course at @jhuclsp , I am leading a class on recent developments in "self-supervised models." Here is the list of the papers and slides we cover: Would love to hear Twitter's suggestions for additional exciting developments to discuss!🤗

6

70

386

Daniel Khashabi 🕊️

@DanielKhashabi

1 year

For anyone interested, we've updated this slide deck on instruction-tuning/RLHF for my class! - PDF: - PPT: Would love to hear your feedback!

Jesse Mu

@jayelmnop

1 year

Since prompting, instruction tuning, RLHF, ChatGPT etc are such new and fast-moving topics, I haven't seen many university course lectures covering this content. So we made some new slides for this year's CS224n: NLP w/ Deep Learning course at @Stanford !

20

292

2K

0

56

203

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

Today we are releasing GENIE🧞, a human-in-loop leaderboard for the evaluation of text generation tasks! We view this as a step forward towards streamlining human evaluation and making it more accessible. #NLP

GENIE: Toward Reproducible and Standardized Human Evaluation for...

While often assumed a gold standard, effective human evaluation of text generation remains an important, open area for research. We revisit this problem with a focus on producing consistent...

arxiv.org

3

44

192

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

It is concerning that an increasing number of research papers base the core of their studies/findings on the new GPT3 models (especially 'davinci-002'), which we know little about training/tuning. How can we do scientific research on these murky foundations?

7

12

177

Daniel Khashabi 🕊️

@DanielKhashabi

1 year

Self-supervised models are a must-know for CS undergrads entering the job market. This semester I taught my first undergrad/MS course on these models, exploring their impact.The course content (slides/assignments) is online for those interested:

4

48

163

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Congrats to whoever graduating this year!! 👏 Please take a few minutes to read and share my piece for @dailypenn on why "I am mourning at graduation", thanks to politics: @UndoFamilyBan #travelban

Guest Column by Daniel Khashabi | Thanks to politics, I am mourning at graduation

In just a few days, the class of 2019 is going to graduate. But many individuals, like me, will walk across the stage without any parents present at the ceremony, thanks to President Trump.

www.thedp.com

7

51

120

Daniel Khashabi 🕊️

@DanielKhashabi

1 year

Overheard someone say GPT-4 is "the end of NLP and CV". That is as absurd as suggesting that iPhone's first release in 2007 marked the end of phone technology. This is not "an end" but rather the beginning of a new era of technological advancements and applications.

3

10

113

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

Excited that our big collaborative effort, "ParsiNLU: A Suite of Language Understanding Challenges for Persian" will appear in TACL'21! If you're working on multilingual/cross-lingual NLP, give it a look! Paper:

ParsiNLU: A Suite of Language Understanding Challenges for Persian

Despite the progress made in recent years in addressing natural language understanding (NLU) challenges, the majority of this progress remains to be concentrated on resource-rich languages like...

arxiv.org

3

37

114

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

📢 GooAQ 🥑: 3 million questions/answers, with a variety of answer types! Draft: Data: 🚨Spoiler alert:🚨 we observe that short- vs long-answer questions behave differently!

GitHub - allenai/gooaq: Question-answers, collected from Google

Question-answers, collected from Google . Contribute to allenai/gooaq development by creating an account on GitHub.

github.com

1

24

108

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

📢 New Preprint📢 Is it possible to provide meaningful textual interpretations of continuous prompts? Find out in our new preprint! Joint work with @sewon__min , @Lianhuiq @sameer_ @wellecks @HannaHajishirzi @tusharkhot @YejinChoinka and others.

Prompt Waywardness: The Curious Case of Discretized Interpretation...

Fine-tuning continuous prompts for target tasks has recently emerged as a compact alternative to full model fine-tuning. Motivated by these promising results, we investigate the feasibility of...

arxiv.org

6

22

108

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

Excited to highlight our work, "Cross-Task Generalization via Language Instructions" TLDR; Language instructions improve generalization to "unseen" tasks. The gains increase w/ more observed tasks. Joint w/ @Swarooprm7 @cbaral @HannaHajishirzi

5

29

102

Daniel Khashabi 🕊️

@DanielKhashabi

4 months

PhD admissions are underway! Showcasing 3 standout students blazing trails in AI/NLP/LLMs (1/4).

3

16

96

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Excited about [re]joining Allen AI @allen_ai ! Over the past few years, AI2 has been at the forefront of key developments in AI/NLP & it's an honor to be part of this vibrant community.

10

2

97

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Hello NLPverse, Want to add a little theoretical spice 🌶️ to your NLP reading list? Check out our theoretical study of multi-step reasoning in the context of language problems; it draws ideas from random graphs & probability theory. 🔥🔥 #NLProc

4

20

90

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

CALL FOR CONTRIBUTIONS: We are soliciting contributions of tasks to a collaborative benchmark of tasks and their natural language instructions/definition. 🚩 Blog: 🤖 Github repo: #NLProc #ArtificialIntelligence

GitHub - allenai/natural-instructions: Expanding natural instructions

Expanding natural instructions . Contribute to allenai/natural-instructions development by creating an account on GitHub.

github.com

2

27

74

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

Text Modular Networks: Learning to Decompose Tasks in the Language of Existing Models (to appear in NAACL'21) 💥 Paper: 🖥️ Demo: 🚧 Code:

1

13

72

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

🥳 New dataset release! 🥳 ARC-DA dataset, a direct-answer (“open response”, “freeform”) QA dataset for elementary-school science domain. Paper: Dataset: Joint work w/ Aristo team at @allen_ai .

ARC Direct Answer Questions Dataset — Allen Institute for AI

A dataset of 2,985 grade-school level, direct-answer science questions derived from the ARC multiple-choice question set.

allenai.org

0

17

65

Daniel Khashabi 🕊️

@DanielKhashabi

10 months

Honored and excited to receive this award for using AI to enhance K12 education! 😊

JHU Computer Science

@JHUCompSci

10 months

Proposals by CS faculty @chienming_huang , @DanielKhashabi , and @ben_vandurme have been selected by the Office of the Provost to receive 2023 DELTA Awards. Learn more about how they plan to leverage the power of #AI in educational settings:

0

5

29

5

1

65

Daniel Khashabi 🕊️

@DanielKhashabi

4 years

Hola #NLProc family, UnifiedQA is now available on @huggingface 's model-hub! 🤗 #Vote2020 #VoteHimOut

Hanna Hajishirzi

@HannaHajishirzi

4 years

New work at @emnlp2020 Findings. UnifiedQA that achieves SOTA in many recent QA datasets by multitask learning and moving beyond format. With @DanielKhashabi @sewon__min and Aristo team at @allen_ai

0

5

42

1

14

60

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

While there are many interesting aspects to the recent "prompting" literature, the fact so much research/energy is spent on effective ways to "engineer" them is indicative of models' brittle comprehension -- hence, not so great news.

1

4

58

Daniel Khashabi 🕊️

@DanielKhashabi

4 years

#NLProc post: Check out our recent-ish work on counter-factual data augmentation (to appear in EMNLP): Natural Perturbation for Robust Question Answering In collaboration w/ Tushar Khot and Ashish Sabharwal.

1

2

48

Daniel Khashabi 🕊️

@DanielKhashabi

1 year

Grateful to Amazon for supporting our research! #AmazonResearchAwards @jhuclsp @JHUCompSci @HopkinsEngineer

Amazon Science

@AmazonScience

1 year

Representing 54 universities in 14 countries, the #AmazonResearchAwards recipients will have access to more than 300 Amazon public datasets, along with AWS AI/ML services and tools. Congrats to the fall 2022 awardees!

1

16

70

8

7

48

Daniel Khashabi 🕊️

@DanielKhashabi

4 years

📢 New work, to appear in @emnlp2020 -Findings 📢 UnifiedQA: Crossing Format Boundaries With a Single QA System Joint with @sewon__min @HannaHajishirzi and other collaborators at @allen_ai .

UnifiedQA: Crossing Format Boundaries With a Single QA System

Question answering (QA) tasks have been posed using a variety of formats, such as extractive span selection, multiple choice, etc. This has led to format-specialized models, and even to an...

arxiv.org

4

5

43

Daniel Khashabi 🕊️

@DanielKhashabi

9 months

Excited about the burgeoning innovations in aligning LLMs to follow"instructions"? Consider participating in the Instruction Tuning & Instruction Following Workshop at #NeurIPS2023 Co-organized w/ @qinyuan_ye @yizhongwyz @ShayneRedford @Francis_YAO_

0

9

41

Daniel Khashabi 🕊️

@DanielKhashabi

10 months

Thrilled by JHU's ongoing commitment to AI. Notably, Hopkins plans to recruit a substantial number of faculty members in the coming few years. Come join us!

Johns Hopkins makes major investment in the power, promise of data science and artificial intelli...

A new institute will bring together experts from a wide range of disciplines to capitalize on the rapidly emerging potential of data to fuel discovery

hub.jhu.edu

2

7

40

Daniel Khashabi 🕊️

@DanielKhashabi

4 years

📢📢 #NLProc post 📢📢 Ever wondered how NLP models view different countries/nationalities? 🤔 Check out this demo of our recent work (to appear in EMNLP-Findings): Join work w/ @tao__li @tusharkhot A. Sabharwal @viveksrikumar

Tao Li

@tao__li

4 years

We present UnQover, a framework to evaluate stereotyping biases in QA models. Tricky to do it since they are often covered up by reasoning errors. Paper: Code: And a beautiful demo:

0

2

13

1

8

35

Daniel Khashabi 🕊️

@DanielKhashabi

1 year

Drago was a kind and enthusiastic person and his passing is a great loss to the community. During my PhD, he invited me to his lab and introduced me to his students, creating opportunities for exchanges/collaboration. It's the little things like this that have an enormous impact!

William Wang

@WilliamWangNLP

1 year

I was deeply saddened to learn of the passing of Prof. Drago Radev. Anyone who interacted with Drago knew he was THE KINDEST PERSON IN THE ENTIRE #NLProc Community. 🕯️🙏 1/N

1

11

148

0

1

33

Daniel Khashabi 🕊️

@DanielKhashabi

2 months

A work led by @jeff_cheng_77 shows that LLMs' knowledge tends to be stale compared to the claimed pre-training cutoff date. As the figure below shows, the effective cutoff of LLMs can be months or even years (!!) earlier than the date claimed by their designers! 🤯🤯

1

3

33

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

Last week we had @_jasonwei in our class to tell us his latest findings on prompting & scaling self-supervised models. Give it a listen:

Emergence and reasoning in large language models - Jason Wei (Google)

Presented as part of CSCI 601.771: Self-supervised Statistical Models: https://self-supervised.cs.jhu.edu/

www.youtube.com

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

For my first course at @jhuclsp , I am leading a class on recent developments in "self-supervised models." Here is the list of the papers and slides we cover: Would love to hear Twitter's suggestions for additional exciting developments to discuss!🤗

6

70

386

0

6

31

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Ever wondered about scaling up NLP technologies to address issues that don't have a simple/single answer? Take a look at our recent work on discovering "diverse perspectives" about controversial issues (accepted to NAACL'19)!!🌈🏳️‍🌈 #NLProc @NAACLHLT @naacl

1

4

30

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

I'm giving a talk at @stanfordnlp this Thursday. If you're around, come hangout with me after the talk.

0

9

29

Daniel Khashabi 🕊️

@DanielKhashabi

1 year

It's been over two years since we put out GENIE! Since then, we have done around ~85 rounds of human evaluation. Today we are releasing all the human annotations for GENIE to benefit the broader research community. DATA:

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

Today we are releasing GENIE🧞, a human-in-loop leaderboard for the evaluation of text generation tasks! We view this as a step forward towards streamlining human evaluation and making it more accessible. #NLP

3

44

192

1

4

29

Daniel Khashabi 🕊️

@DanielKhashabi

7 months

How can we make LLMs robust to noise in the training data? 🤔 We propose "error norm truncation", a modified training objective that suppresses noisy data, improves model accuracy, and speeds up the convergence! Paper:

Error Norm Truncation: Robust Training in the Presence of Data...

Text generation models are notoriously vulnerable to errors in the training data. With the wide-spread availability of massive amounts of web-crawled data becoming more commonplace, how can we...

arxiv.org

Tianjian Li

@tli104

8 months

(1/5) The standard MLE objective is notoriously vulnerable to noise! How can we make LLMs robust to noise in the training data? 🤔 We propose Error Norm Truncation (ENT), a modified training objective that ignores noisy tokens in the training corpus. 📰:

1

8

14

1

4

28

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

My summary of the major highlights of "natural language understanding," over the past 60 years. Items are color-coded based on their contribution-type. CPU/GPU speed on the side, to provide perspectives about the role of the computational resources. From:

1

4

29

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

StrategyQA dataset now has a leaderboard! 🎉

AI2 Leaderboard

The AI2 Leaderboard platform hosts public leaderboards for a variety of AI challenge tasks across multiple research domains.

leaderboard.allenai.org

Mor Geva

@megamor2

3 years

We present StrategyQA, a question answering benchmark with *implicit* reasoning strategies, accepted to TACL, 2021. Dataset --> Paper --> With @DanielKhashabi @EladSegal @tusharkhot @dannydanr @JonathanBerant

3

34

152

1

5

29

Daniel Khashabi 🕊️

@DanielKhashabi

1 year

Venugopal et al. 2011 () is a pioneering paper on "watermarking" generative models that was two decades ahead of its time! Most recent papers on text/data watermarking use techniques that are quite similar to this old-ish work, alas they don't mention it.🤦

2

1

28

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

Want to solve reading comprehension? Solve MultiRC! (Multi-Sentence Reading Comprehension) #NLProc #ArtificialIntelligence @cogcomp

1

6

28

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

Many are excited about continued gains with over-parameterized models. I was recently surprised (and delighted) to learn about pioneering works from **~20 years ago** that show the benefits of the increased parameter count. Here are two that caught my eye:

1

26

Daniel Khashabi 🕊️

@DanielKhashabi

7 years

@stanfordnlp 's TrueCase annotator is one of the most underrated and less-appreciated tools in NLP.

CoreNLP

High-performance human language analysis tools, now with native deep learning modules in Python, available in many human languages.

stanfordnlp.github.io

0

5

26

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

We studied this problem recently and the results are quite troubling: Just scratching the surface; we (the research community) have a long way to go.

UnQover Demo

UnQover is a general framework to probe and quantify social biases (e.g., bias against ethnic/racial groups, nationalities, genders, etc) in QA models.

unqover.apps.allenai.org

Ani Nenkova

@ani_nenkova

3 years

Contemporary language models encode all sorts of stereotypes expressed in the data used for their training. There is no algorithmic way to list all learned stereotypes and there is no effective way to fix these. New technology has to be developed with this in mind.

3

12

125

1

6

26

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

(1/4) Check out ZOE, Zero-Shot (no manually annotated data for typing) and Open (working on a wide range of target labels) entity typing. Paper: Code: Joint work with @BenZhou96 @chentsetsai @dannydanr @cogcomp #NLProc

2

7

26

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

اگه به #پردازش_زبان_طبیعی (بخصوص روی فارسی) علاقمندید، به کار ما نگاهی بندازید!‌

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

Excited that our big collaborative effort, "ParsiNLU: A Suite of Language Understanding Challenges for Persian" will appear in TACL'21! If you're working on multilingual/cross-lingual NLP, give it a look! Paper:

3

37

114

0

1

25

Daniel Khashabi 🕊️

@DanielKhashabi

7 months

Do #LLMs learn "In-context" by "Gradient Descent"? See our skeptical take below 👇

0

5

24

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

Must-watch talk by @GaryMarcus

2

8

22

Daniel Khashabi 🕊️

@DanielKhashabi

8 years

Shameful advertisement: Cool structured learning library in Java: #machinelearning #artificialIntelligence #NLProc

1

13

20

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

This is basically an empirical take on @brianchristian 's recent book: "The Alignment Problem: How Can Machines Learn Human Values?" So far, unfortunately, the answer is a "no".

Jieyu Zhao@ICLR2024

@jieyuzhao11

3 years

Can we intervene in a model’s behavior by natural languages? Check our #ACL2021 Findings “Ethical-Advice Taker: Do Language Models Understand Natural Language Interventions?” (). w/ @DanielKhashabi , Tushar Khot, Ashish Sabharwal, and @kaiwei_chang . 1/n

2

16

84

2

4

21

Daniel Khashabi 🕊️

@DanielKhashabi

4 years

Mined a several thousands of search queries related to coronavirus/covid-19. Here is the data: Need your help here: how can we use these queries to addresses a real challenge we now face?

1

21

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

Data and baselines: Joint work w/ @armancohan @pedramhosseini @ppezeshkpour @malihealikhani @MarziehBitaab @faeze_brh @mgheini @karimirabeeh @omidmemari @erfannoury @shahab_raji @sepid_s @saber_sh93 @ali_tazarv @yyaghoobzadeh

GitHub - persiannlp/parsinlu: A comprehensive suite of high-level NLP tasks for Persian language

A comprehensive suite of high-level NLP tasks for Persian language - persiannlp/parsinlu

github.com

0

7

21

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

کتاب یادگیری ماشین – به فارسی @tommmitchell 's Machine Learning, in Persian: Courtesy of my dear friend, Mohammad Nokhbeh Zaeem.

0

2

19

Daniel Khashabi 🕊️

@DanielKhashabi

7 months

Looking for effective prompts without breaking the bank?💰 (1) Prompts with flatter loss [surrogate] minima generalize better. (2) LLMs' flatness can be efficiently approximated via a surrogate function (little/no labeled data). Paper:

0

2

20

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Excited about new work with the up-and-coming researcher, @BenZhou96 !! The dataset leaderboard is available now:

AI2 Leaderboard

The AI2 Leaderboard platform hosts public leaderboards for a variety of AI challenge tasks across multiple research domains.

leaderboard.allenai.org

Ben Zhou

@BenZhou96

5 years

Check out our new #emnlp2019 paper where we studied temporal commonsense: . We collected a QA dataset MC-TACO🌮(leaderboard coming soon) and showed that it's a new challenge to existing systems. Co-author with @DanielKhashabi , Qiang Ning and Dan Roth.

0

2

14

0

20

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Trump might let it go; @Grammarly won’t.

2

0

18

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

Absolutely amazed by the massive amount of contributions that we have received from the community! 🥳🥳 One more week to go (mid October), if you'd like to join the effort! #NLProc

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

CALL FOR CONTRIBUTIONS: We are soliciting contributions of tasks to a collaborative benchmark of tasks and their natural language instructions/definition. 🚩 Blog: 🤖 Github repo: #NLProc #ArtificialIntelligence

2

27

74

1

2

19

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Feel opinionated about certain a topic and want to see how other people think about it? Try our demo and check if it helps you see the alternative "perspectives." Paper: Video: Demo: @cogcomp @ccb

PerspectroScope: A Window to the World of Diverse Perspectives

Read more in the following paper: PerspectroScope: A Window to the World of Diverse Perspectives, ACL - Demos, 2019.

www.youtube.com

0

3

17

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

If you really care about global warming, **divest** from bitcoin (or similar forms of cryptocurrency).

Bitcoin emissions alone could push global warming above 2°C

Nature Climate Change - Bitcoin is a power-hungry cryptocurrency that is increasingly used as an investment and payment system. Here we show that projected Bitcoin usage, should it follow the rate...

www.nature.com

0

3

17

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

Check out this excellent work by @kel_lu and many great collaborators @allen_ai & @uwnlp ! Spoiler alert: temporal adaptation (further pre-training) is nowhere enough to solve the temporal drift of pre-trained language-models on downstream tasks.

kellu

@kel__lu

3 years

In our new paper, we investigate how temporal misalignment, when a model is trained on data from one time period but tested or deployed on data from another, affects NLP models across a variety of tasks and domains. (1/n)

2

6

52

0

17

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

If you're at #EMNLP , make sure to stop by @BenZhou96 's poster to hear about his work on Zero-Shot + Open entity typing Grand Hall, 09:00 – 10:30. #NLProc

0

4

16

Daniel Khashabi 🕊️

@DanielKhashabi

1 year

In the '70s, many correlated a computer's size with its computational strength. Obviously, that is no longer the case, as we each carry powerful computers in our pockets. Will our attitude towards "large self-supervised models" evolve similarly? Only time will tell!

0

2

16

Daniel Khashabi 🕊️

@DanielKhashabi

4 months

Lingfeng Shen @Lingfeng_nlp () is a powerhouse, and his CV speaks for itself. Checkout Lingfeng's work that will appear in ICLR '14:

The Trickle-down Impact of Reward (In-)consistency on RLHF

Standard practice within Reinforcement Learning from Human Feedback (RLHF) involves optimizing against a Reward Model (RM), which itself is trained to reflect human preferences for desirable...

arxiv.org

1

3

16

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

I will be at @NAACLHLT @naacl presenting our poster (tomorrow June 2, 10:30a). If you're around, stop by! :)

0

2

16

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

To be clear, I am not against using GPT-3; I am against *only* studying GPT-3 and hence, ignoring the generality of findings on other models for which we have more clarity.

1

0

16

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

"UPenn's Department of Philosophy will not require Ph.D. program applicants to submit GRE scores this year." I look forward to hearing similar changes in other schools and departments, to facilitate the admissions process for those who can't afford it. @umphilosophy @weisbergm

0

2

16

Daniel Khashabi 🕊️

@DanielKhashabi

4 months

Tianjian Li @tli104 () is making waves. He led a project into the spotlight at ICLR '14 (top 5% of submissions) nearly single-handedly.

Error Norm Truncation: Robust Training in the Presence of Data...

Text generation models are notoriously vulnerable to errors in the training data. With the wide-spread availability of massive amounts of web-crawled data becoming more commonplace, how can we...

arxiv.org

1

14

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

Last but not least, I am thankful to @etzioni for cultivating such a vibrant and flourishing community of talented researchers at @allen_ai .

1

14

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

@zehavoc @seb_ruder @DeepIndaba @_aylien Ya, the figure is misleading; just to add to your point: Early 2000s: Introduction of FrameNet. Early 2000s: CoNLL shared tasks which helped significant progress (e.g. in NER). 2001: CRFs 2002: BLEU score, let MT systems scale up. 2002: Early PropBank ~2002: Topic Models

2

0

14

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

Blanket sanctions hurt ordinary people (blocking Iran's access to life-saving medicines, passenger planes, etc) - if you're rejoicing sanctions, remember that you are depriving +80M people of normal life. #nosanctionnowar

0

1

13

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

@zehavoc @seb_ruder @DeepIndaba @_aylien 2004: RTE 2004-7: Framenet Parsers. 2004: Early relation extraction models 2002-2006: Early Semantic Role Labelers. 2006: Ontonotes ~2007: Open IE ~2008: Distant supervision ~2010: Wikification and entity-linking ~2010: NELL ... (and many others missing)

2

0

13

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

It's okay if in short term these analyses inform our understanding of models' weaknesses/strengths, though I am more excited about a future where models are robust/competent against a *variety* of natural lang commands/instructions (and hopefully less or no "prompt engineering").

0

13

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

Summary of tools included in CogCompNLP pipeline, @stanfordnlp CoreNLP, @spacy_io , @ApacheOpennlp , NLTK and TextBLOB -- Github: -- Demo:

1

9

13

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

It took a few years, but eventually, democracy worked! Thanks, @POTUS ! #MuslimBan

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Congrats to whoever graduating this year!! 👏 Please take a few minutes to read and share my piece for @dailypenn on why "I am mourning at graduation", thanks to politics: @UndoFamilyBan #travelban

7

51

120

0

13

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

CogCompNLP: Summary of annotators vs @stanfordnlp , @spacy_io , @ApacheOpennlp , NLTK -- Github: … -- Demo: @lrec2018 #NLProc

2

12

13

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

If you're attending NAACL, stop by our poster tomorrow morning! Draft: More: @naacl #naacl2019 🏳️‍🌈

2

13

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

While my title may change, I know I will remain a lifelong student. I am looking forward to learning and growing alongside the many young bright minds I will meet at JHU.

1

13

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

Imagine reading a paper with lots of cool findings based on an obscure model X (rather than GPT3). Would you buy them as general findings? Ever wonder if their findings might be specific to their model?

1

0

13

Daniel Khashabi 🕊️

@DanielKhashabi

7 years

Transliteration into about 100 languages, by @mayhewsw #NLP

0

3

13

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

(4/4) Zoe and Elmo are finally back together! :) (pictured: Elmo and Zoe)

0

12

Daniel Khashabi 🕊️

@DanielKhashabi

1 month

As a @Penn alum and an academic myself, I am disheartened to hear about this decision. This isn't about choosing sides between Palestinians or Israelis. Instead, it's about the fundamental role universities should fulfill within our society.

The Daily Pennsylvanian

@dailypenn

1 month

BREAKING: Penn Students Against the Occupation of Palestine’s status as a registered student group has been revoked "effective immediately," a University spokesperson told The Daily Pennsylvanian.

6

7

17

1

3

12

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

AI research needs transparency and openness for us to continue building upon each other's shoulders (CC: @OpenAI , @sama ).

1

0

11

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

"This has significant business consequences, What I can say — with complete confidence — is you are going to see a whole new generation of products, some from start-ups, some from the big companies." Oren Etzioni @etzioni comments on a major AI milestone, at @allen_ai !

Oren Etzioni

@etzioni

5 years

. @allen_ai 's Aristo aces 8th and even 12-Grade Science Tests what do these results tell us about NLP? About reasoning? thought-provoking article by @cademetz

5

59

135

1

11

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

I look forward to extensive collaborations at @jhuclsp @jhucompsci and I am excited to play my part in helping usher in the future of equitable, transparent, and reliable AI. The field has come a long way, and the best is yet to come!

2

0

11

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

Excited to try my brand new @google -home, tried asking it to play a @chaartaar music: "Ok Google! Play a Chartaar song"; but it kept confusing it for "charter", "chart a", "chart". (but wait, someone said #ArtificialIntelligence is taking over the world?)

1

0

11

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

Westerners awarding a poorly-produced movie that aligns well with their pre-existing, incomplete perception about a different culture/society ... yikes! We're stuck in our loop of biases and stereotypes.

International Emmy Awards

@iemmys

2 years

The International Emmy for Drama Series goes to “Tehran” produced by Donna and Shula Productions / Paper Plane Productions! #Israel #iemmyWIN

41

115

600

1

10

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

I am greatly indebted to all my collaborators who made this possible, particularly my *amazing* mentors @DanRothNLP , @YejinChoinka , @HannaHajishirzi and Ashish Sabharwal.

1

10

Daniel Khashabi 🕊️

@DanielKhashabi

7 years

Great work by @ybisk and @aliceylai , on extending the entailment problem. #NLProc

0

1

10

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

We were lucky to have Anjalie Field ( @anjalie_f ) in our class to tell us about "Social Applications of Pre-trained Language Models". For those interested, here is the recording:

Social Applications of Pre-trained Language Models - Anjalie Field...

The talk was given to CSCI 601.771: Self-supervised Statistical Models: https://self-supervised.cs.jhu.edu/Bio: Anjalie Field is a postdoctoral Fellow at St...

www.youtube.com

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

For my first course at @jhuclsp , I am leading a class on recent developments in "self-supervised models." Here is the list of the papers and slides we cover: Would love to hear Twitter's suggestions for additional exciting developments to discuss!🤗

6

70

386

0

1

10

Daniel Khashabi 🕊️

@DanielKhashabi

8 years

Nice work by @vardi group! Improved approx for counting CNF feasibility space #IJCAI16 #ArtificialIntelligence

0

4

10

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Disappointed that none of the candidates tonight spoke on "the travel ban." #DemDebate #TravelBan

1

10

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

"We All Do Better When We All Do Better"

0

9

Daniel Khashabi 🕊️

@DanielKhashabi

4 years

Thanks @PennSAS !

Jamal Abdi

@jabdi

5 years

NIAC Urges Universities to Extend Admission Deadlines for Iranian Students amidst Internet Shutdown - thank you to everyone who flagged this and is already working with administrators

3

13

54

0

2

9

Daniel Khashabi 🕊️

@DanielKhashabi

4 years

Oh the irony is really something.

Donald J. Trump

@realDonaldTrump

4 years

To the leaders of Iran - DO NOT KILL YOUR PROTESTERS. Thousands have already been killed or imprisoned by you, and the World is watching. More importantly, the USA is watching. Turn your internet back on and let reporters roam free! Stop the killing of your great Iranian people!

27K

50K

245K

1

8

Daniel Khashabi 🕊️

@DanielKhashabi

3 years

... while short answer questions benefit heavily from more labeled data, long answer questions are mostly driven by pre-training of the models. Joint work w/ Amos Ng @tusharkhot Ashish Sabharwal @HannaHajishirzi @ccb

0

2

8

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

Mari Ostendorf: "Despite the recent hype, ASR is not solved yet" #nlpto #NAACL2018

0

1

8

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Folks in NLP/Data-mining interested in information pollution/distortion/manipulation in social media: Here is a nice case study for you, happening *now* in a massive scale. Nice work and visualization by @geoffgolberg #NLProc #datamining #DataScience

0

4

8

Daniel Khashabi 🕊️

@DanielKhashabi

5 years

Once a colony 🇮🇳 has surpassed its colonizer 🇬🇧 in terms of GDP. There is so much to learn here.

1

0

8

Daniel Khashabi 🕊️

@DanielKhashabi

2 years

Lawrence, Giles, and Tsoi (1997) show that models with larger hidden units can lead to consistently better generalization for the face recognition task. In particular, their best generalizing network (shown in fig below) had 364x more parameters (18k params) than training data.

1

8

Daniel Khashabi 🕊️

@DanielKhashabi

6 years

"In the early twenty-first century, the average human is far more likely to die from bingeing at McDonald's than from drought, Ebola or an al-Qaeda attack." (Homo Deus, A Brief History of Tomorrow; Yuval Harrari)

1

8

Daniel Khashabi 🕊️

@DanielKhashabi

4 years

"The continued decline in international student enrollment since the fall of 2016 has cost the US economy $11.8 billion and more than 65,000 jobs, according to estimates from NAFSA (Association of International Educators)"

0

8

Daniel Khashabi 🕊️

@DanielKhashabi

8 years

I haven't used @stanfordnlp 's tokenRegex, but the description looks great; I wonder how fast/scalable it is tho...

1

3

8