Anna Rogers 🇺🇦🇪🇺 is looking for postdocs! @annargrs Twitter profile | Pikagi

Pikagi

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

@annargrs

8,926

Followers

864

Following

299

Media

5,269

Statuses

Associate professor @ITUkbh : LLM interpretability, generalization, AI & society. Co-editor-in-chief @ACLRollingReview

Copenhagen, Denmark

https://t.co/fiprSJYn52

Joined September 2018

Don't wanna be here? Send us removal request.

Pinned Tweet

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

9 months

Excited to work with @SashaMTL on this! A new position paper discussing what we even mean by "LLM", and some of the common claims/assumptions about them: - LLMs are robust - LLMs are SOTA - Scale is all you need - LLMs exhibit emergent properties

@SashaMTL

Sasha Luccioni, PhD 🦋🌎✨🤗

9 months

New position paper just dropped! 💣 In it, @annargrs take a critical look at LLMs, from their definition to their functionality and influence on the field of NLP. A brief thread of our arguments 🧵:

Tweet media one

3

37

153

3

10

86

Last Seen Profiles

@VincentBanks128

@xiaomelonz

@HaydenHewitt

@hakuoh_soldiers

@GerryAdamsSF

@ShaneSaint

@yoshiebbc

@Nils_Wollny

@BS4chan

@skinworms

@GregorBrand

@Baladria

@keelgaa

@PetSatoshi

@thehollowhorn

@flo_wershower

@Tom_Suozzi

@itsmarlamaples

@EDHijinks

@portalkatyperry

@jiheown

@_dyats

@GODDESS_DAY_

@matichonbooks

@LoveranOfficial

@scrawberyy

@julie_renbe

@miyumiyumiyua79

@TexansforRFKJr

@renaper_ar

@TravelTeach31

@AngieNOLA

@_iamprodigy

@_UzbekFootball

@cellyruMFR

@CoachWesCope

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

❌"NATO expansion to the East" ✅ "Eastern Europe leaving the ex-empire" The NATO frame makes it about Russia vs NATO, denying agency to Eastern European countries. The fact is: they've all been jumping the Soviet ship for decades, at the first opportunity. It's kinda clear why.

22

561

3K

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

Have you been drowning in BERT papers? We have. So... the first ever... Primer in BERTology is out! We survey over 40 papers on BERT's linguistic knowledge, compression, architecture tweaks, multilinguality, and more! With OlgaKovaleva & @arumshisky .

6

315

1K

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

📢 tech & ethics folks: I've collected 🧵 of pointers to official statements and discussions on what the big tech has done/should do for 🇺🇦 war. Again, this is not just about 🇺🇦: what we do now sets precedent for the future. Comments, updates & corrections welcome! /1

17

326

838

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

5 years

Here's a summary post on problems with huge models that dominate #NLProc these days. I put together several different discussion threads with/by @yoavgo , @jaseweston , @sleepinyourhat , @bkbrd , @alex_conneau , @SeeTedTalk .

How the Transformers broke NLP leaderboards

With the huge Transformer-based models such as BERT, GPT-2, and XLNet, are we losing track of how the state-of-the-art performance is achieved?

hackingsemantics.xyz

8

155

440

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 months

🥳🍾 It's official - I'm a tenured associate professor! This job is incredible luck and privilege, and @ITUkbh is an amazing place to work at!!! More PhD student and postdoc positions will be announced soon.

42

11

420

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

A belated blog post for our BERTology EMNLP paper (by Olga Kovaleva, Alexey Romanov, yours truly and @arumshisky ). My favorite experiment in this work is showing that for most GLUE tasks BERT works pretty well even *without pre-training*!

Tweet card media

The Dark Secrets of BERT

BERT and its Transformer-based cousins are still ahead on all NLP leaderboards. But how much do they actually understand about language?

text-machine-lab.github.io

4

126

411

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

Life update: I'm now an assistant professor at @CPH_SODAS ! Very grateful to be in the right place to learn from all those psychologists, anthropologists, social & political scientists - all the people who actually know what we're baking into NLP models, apart from language.

64

14

397

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

5 months

@chrmanning at #EMNLP2023 to #NLProc PhD students, who are having an existential crisis over LLMs: Aeronautics students do not build Boeings for their PhD theses. They do smaller models - and still make meaningful contributions. There's plenty of such opportunities for us too.

Tweet media one

Tweet media two

5

76

394

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

🎈 #NLPaperAlert : Changing the World 🌍 by Changing the Data 🗃 A soul-searching piece that made it to ACL 2021: - how NLP resources affect the world - what does it even mean to 'work in NLP' - how we can make better use of our subcommunities. /1

Tweet card media

Changing the World by Changing the Data

NLP community is currently investing a lot more research and resources into development of deep learning models than training data. While we have made a lot of progress, it is now clear that our...

8

87

388

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

Will reviewers now ask "where's GPT4?" We propose a #BenderRule -like rule: That which is not open & reproducible* can't be a requisite baseline w @b_niranjan @LeonDerczynski @JesseDodge @alkoller @SashaMTL @MaartenSap @royschwartzNLP @nlpnoah @strubell

Tweet card media

Closed AI Models Make Bad Baselines

Will GPT-4 become a universally expected baseline in NLP research, like BERT in its time? Basic scientific methodology demands otherwise.

hackingsemantics.xyz

10

101

370

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

> ChatGPT was more empathetic than humans 🤦 No it wasn't! It produced text that humans, unaware of its origin, interpreted as empathetic. Dear journalists, I know that tweets are short, but these language shortcuts do matter in how public perceives LLMs. It's not just jargon.

@TamarHaspel

Tamar Haspel

1 year

Researchers pittted ChatGPT against genuine doctors, using actual patient questions. ChatGPT was more accurate and more empathetic. By a lot. HT @adamcifu

Tweet media one

400

1K

9K

26

66

365

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

Personal news: I'm moving to Copenhagen! Excited to join the interdisciplinary team of political scientists, psychologists, anthropologists, economists, and computer scientists at @CPH_SODAS . I'll be working on both core and social NLP topics.

23

6

292

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

New paper📜: What Can We Do to Improve Peer Review in NLP? with @IAugenstein TLDR: In its current form, peer review is a poorly defined task with apples-to-oranges comparisons and unrealistic expectations. /1

Tweet media one

7

70

275

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

#NLPaperAlert : QA Dataset Explosion!🔥 A survey of 200+ QA/RC datasets proposing a taxonomy of formats & reasoning skills. Also in the bag: modalities, conversational QA, domains & beyond-English data. Honored to work on this with @nlpmattg & @IAugenstein

Tweet media one

10

70

263

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

A highlight from the amazing @ev_fedorenko 's keynote: The language system in the human brain at #EMNLP2021 It doesn't really make sense to criticize neural LMs for not being able to reason, because the human language system is also separate from other cognitive abilities. /1

Tweet media one

4

45

258

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

#NLPaperAlert Our Primer in BERTology is finally up in TACL! A collection of insights from over 150 papers: how BERT works, suggested improvements, open questions and starting points for new research. Joint work with Olga Kovaleva and @arumshisky

Tweet card media

A Primer in BERTology: What We Know About How BERT Works

Abstract. Transformer-based models have pushed state of the art in many areas of NLP, but our understanding of what is behind their success is still limited. This paper is the first survey of over...

1

53

243

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

New blog post: BERT Busters👻: outlier dimensions that disrupt Transformers TLDR: We found the Achilles' heel of BERT&co! Disabling just <0.0001% of outlier layer output weights drops model quality this much: /1

Tweet media one

5

42

233

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 months

Yay, I got my first grant! 🙏 @DFF_raad I'll be looking for a postdoc to work on an ambitious generalization benchmark, in collaboration with @tallinzen . If you'd like to spend a year in beautiful Copenhagen - please reach out! (email preferred) Could start as early as February.

13

16

228

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

10 months

*ACL deadlines for 2024 (all through ARR)

Tweet media one

2

64

210

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

📢 Are you interested in both AI and how it interacts with human society? Would you like to do a truly interdisciplinary PhD - not just DL, but also actually learning about the subject matter? @CPH_SODAS can host projects co-supervised by real social scientists + yours truly! /1

5

57

183

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

Ok this is official: yours truly will be one of PC chairs for @aclmeeting 2023! Thrilled to work on this with @boydgraber @chokkanorg & Yang Liu. No pressure... we just need to organize one of the top NLP events... how hard could it be😅

@eltimster

Tim Baldwin

2 years

@ani_nenkova @aclmeeting General Chair: Yang Liu; PC Chairs: @annargrs @boydgraber and @chokkanorg

1

0

8

8

9

180

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

🥁 The 2022 Insights best paper award 🏆 goes to... On the Impact of Data Augmentation on Downstream Performance in NLP Itsuki Okimura @machelreid @mkt_kwn @ymatsuo PDF: To be presented in the thematic session 3 at 12pm Thanks to @GoogleAI sponsorship!

Tweet media one

2

34

179

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

Personal news: I'll soon start as an assistant professor at @ITUkbh , while remaining a @CPH_SODAS fellow. Very happy to be able to stay in Denmark!

20

3

178

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

Our blog post on why "closed" models like ChatGPT shouldn't be considered requisite baselines, no matter the hype, made it to @TDataScience ! With @b_niranjan @LeonDerczynski @JesseDodge @alkoller @SashaMTL @MaartenSap @royschwartzNLP @nlpnoah @strubell

Tweet card media

Closed AI Models Make Bad Baselines

This post was authored by Anna Rogers, with much invaluable help and feedback from Niranjan Balasubramanian, Leon Derczynski, Jesse Dodge…

towardsdatascience.com

4

29

172

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

Continuing to learn from the blogging masters at @rethinkmlpapers : @lilianweng shares her workflow, which is based on her own daily reading. She takes notes while reading, and a blog posts emerges when there is enough material around some topic.

Tweet media one

1

25

163

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

The #ACL2023 CFP is up at the conference website! Honored to work on this with @boydgraber , @chokkanorg and Yang Liu. Submissions: ARR + direct ARR deadline: December 15 (yes, it's changed) Softconf deadline: January 20 Special theme: Reality check 😎

Main Conference

Official website for the 61st Annual Meeting of the Association for Computational Linguistics

2023.aclweb.org

4

56

161

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

A few highlights from @nlpnoah 's talk at : @nlpnoah : NLP research and practice ask fundamentally different questions /1

Tweet media one

3

41

165

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

Dear tech & ethics people: I've done some research on Roskomnadzor: the "ministry of truth" that created the alternative propaganda reality most Russians seem to inhabit. Some truly fascinating stuff. Here's what I learned 🧵 As always, comments/corrections welcome. /1

8

49

155

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

This is your reminder that in 2020 your failed experiments don't have to disappear. There's still time to submit a short paper or a non-archival abstract to @emnlp2020 workshop for Insights from Negative results! /1

Workshop on Insights from Negative Results in NLP

insights-workshop.github.io

1

27

141

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

🇺🇦 update: Ukrainian army is still standing - which is great news. Generally, the longer they manage to repel the initial attacks, the more the Russians are losing the blitzkrieg advantage. And the worse it looks for them internally and externally. /1

2

7

135

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings. Paper by @ethayarajh , #EMNLP2019

3

35

133

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

#Reviewer2 , you've been warned 😀

@emnlp2020

emnlp2020

4 years

Seems like a lot of people are working on their EMNLP reviews at the moment! Please remember that the following are NOT valid primary reasons for rejection, and the authors will be able to appeal to these guidelines in the rebuttal:

Tweet media one

13

156

558

2

14

129

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

5 years

More reviews saying "this is just another dataset / annotation scheme for task A, there's already too many of those". Why does nobody ever complain "this is yet another model for task A"? #emnlp2019 #NLProc

8

17

127

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

Worried about benchmark data contamination? Studying LLM memorization or attribution? @BigscienceW BLOOM 🌸 now has exact & fuzzy search over full training data! with @olapiktus 🏆 @christopher Paulo Villegas @HugoLaurencon @ggdupont @SashaMTL @YJernite /1

Tweet card media

The ROOTS Search Tool: Data Transparency for LLMs

ROOTS is a 1.6TB multilingual text corpus developed for the training of BLOOM, currently the largest language model explicitly accompanied by commensurate data governance efforts. In continuation...

3

29

126

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

10 months

A few highlights from @AlisonGopnik keynote at #ACL2023 . Do watch it, it was amazing! A few highlights in this thread. /1

Tweet media one

1

21

126

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

5 years

New #emnlp2020 workshop: Insights from Negative results from NLP! Organized by Joao Sedoc, @arumshisky and myself, invited talks by @radamihalcea and @byron_c_wallace . Special focus: breaking pre-trained meaning representations.

Workshop on Insights from Negative Results in NLP

insights-workshop.github.io

2

47

126

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

Join Yoshua Bengio in this petition to allow remote conference presentations: "...so that anyone be free to decide... whether the benefits of attending on site outweigh the negative consequences of the trip - both for climate and for family life."

Tweet card media

Can you spare a minute to help Anna Rogers?

Allow remote paper & poster presentations at scientific conferences

2

55

121

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

Thanks to @ReviewAcl for inviting us to write a tutorial for #NLProc reviewers! This is a complete, step-by-step guide, with an updated list of lazy heuristics that should *not* be considered as weaknesses of a submission (such as reject-if-not-SOTA).

Tweet media one

@ReviewAcl

ACLRollingReview

2 years

We have a new tutorial for reviewers from @annargrs and @IAugenstein ! This covers all aspects of writing a review, from preparation before reviewing, to basic principles of reading for review and providing useful comments, as well as common pitfalls.

3

48

137

4

29

121

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

Likewise: 🇺🇦 content alert. I hope everybody follows what happens, bc it's not just Ukraine. The real question is: "Can one mad bully with a nuke tell the whole world what to do?" If global cooperation can't stop this, it's a precedent. And many other regions have wannabe tsars.

@ducha_aiki

Dmytro Mishkin 🇺🇦

2 years

Dear friends. I realize that you followed me for the news in computer vision, image matching, machine learning. Not for political bullshit. But my country is in grave danger. It is not politics, it is life&death. I will continue to write about ML&CV. And also about 🇺🇦.

58

184

4K

3

9

119

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

I hope it's obvious that (a) is a reptiloid-level conspiracy theory. Ver. (b): sure. So did Russia, with a head start. All election candidates try to influence voters, but the result is still voters' choice, right? Eastern Europe saw what both worlds had to offer, and voted. /3

2

7

110

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

A highlight from #EMNLP2021 fascinating keynote by @StevenBird : NLP often comes with a set of assumptions about what are the needs of communities with low-resource languages. But we need to learn what they *actually* need, they may have a completely different epistemology. /1

Tweet media one

3

25

114

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

Tired of paper pdfs? Brainstorm with us about the future of research communication at @rethinkmlpapers ( @iclr_conf Friday)! Talks & panel by David Ha, Terrence Parr @evelynevs @FalaahArifKhan @Hugo_larochelle @jeffbigham @lillian_weng @deviparikh 🧵 Some ideas from the program:

5

24

114

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

📢 @aclmeeting has issued a clarification on the soundness/excitement scores, and released pre-rebuttal score distirbution. We will also follow up with authors, chairs and reviewers by email. /1

Tweet media one

6

34

113

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

🇺🇦 update: lots of great news this morning! * Kyiv is stil standing. As a matter of fact, the mighty Russian army still isn't holding ANY large city in Ukraine, which is getting embarassing. /1

2

3

111

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

THIS HAPPENED!!! 🥳🎉🍾 My GS profile now shows +100 missing citations. I didn't get the email notification for these - so check your profiles, you may be pleasantly surprised! Tears of joy, and deep thanks to @earnmyturns and the Google Scholar team.

@aclanthology

ACL Anthology

4 years

This long-standing issue seems to have been resolved. Thanks to @annargrs for raising this issue, @earnmyturns for raising this issue internally, and also to Google Scholar for the fix!

4

25

135

6

8

109

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

I really enjoyed this episode of #nlphighlights with @earnmyturns . It is about managing industry research teams, but also generally about incentives in research and the need for intellectual diversity. A few highlights from the highlights: /1

Tweet card media

113 - Managing Industry Research Teams, with Fernando Pereira

We invited Fernando Pereira, a VP and Distinguished Engineer at Google, where he leads NLU and ML research, to talk about managing NLP research teams in industry. Topics we discussed include prioritiz

2

26

109

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

#NLPaperAlert 📢: ‘Just What do You Think You’re Doing, Dave?’ A Checklist for Responsible Data Use in NLP (with @eltimster , @Kobotic ): - summary of the principles of data use & where they don't align - proposal for a #NLProc data use checklist, akin to the reproducibility one /1

Tweet media one

3

41

104

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

📢 #NLPaperAlert #COLING2022 Machine Reading, Fast And Slow: When Do Models "Understand" Language? TLDR: instead of claiming broad "language understanding", why don't we define the reasoning expected in specific cases? with @sagnikrayc @IAugenstein /1

Tweet card media

Machine Reading, Fast and Slow: When Do Models...

Two of the most fundamental challenges in Natural Language Understanding (NLU) at present are: (a) how to establish whether deep learning-based models score highly on NLU benchmarks for the...

2

22

103

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

📢 New blog post: the attribution problem with generative AI #NLProc AI & Society TLDR: Some argue that publicly available data is fair game for commercial models bc human text/art also has sources. But unlike models, we know when attribution is due...

Tweet card media

The attribution problem with generative AI

Some argue that any publicly available text/art data is fair game for commercial models because human text/art also has sources. But unlike models, we know when attribution is due…

hackingsemantics.xyz

4

31

100

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

How can you help🇺🇦? 4 ways: 1) Donate (links I checked myself in this🧵) 2) Bug your representatives for sanctions & aid 3) Information hygiene: great guide by @katestarbird 4) It's not "Ukraine crisis" or "operation". It's "war/invasion". /1

1

36

101

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

5 years

Teaching an intro PyTorch tutorial at ESSLLI. Audience is 80% linguists, and I start very confident I know how to talk to former colleagues. A question halfway into the tutorial: - Sorry, what is "GPU"? (Sound of my brain overclocking to reassess the rest of the material)

1

7

99

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

New #NLProc blog post: crusade against reject-if-not-SOTA conference reviews! TLDR: SOTA performance is just one way in which a system could be interesting. Conference guidelines should clearly include others, such as efficiency and generalizability. /1

Tweet card media

Peer review in NLP: reject-if-not-SOTA

Many reviewers at major NLP conferences tend to reject models that fail to beat state-of-the-art. It is a heuristic that is simple, convenient, and wrong.

hackingsemantics.xyz

8

34

100

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

8 months

🤦 AI X-risk is a global priority. Not its climate impact. Not potential for automated inequality or manipulation of populations. Not chaos from current LLMs with API access to stuff or writing bad code for production. Not required revamp of education with no extra budget.

@EU_Commission

European Commission

8 months

Mitigating the risk of extinction from AI should be a global priority. And Europe should lead the way, building a new global AI framework built on three pillars: guardrails, governance and guiding innovation ↓

Tweet media one

435

485

2K

5

19

99

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 months

📢 #NLProc job alert: I'm looking for a #compling postdoc to work on an ambitious generalization benchmark, in collaboration with @tallinzen . Could start as early as April. If you'd like to spend a year in beautiful Copenhagen 🇩🇰 - please reach out! (arog @itu .dk) RT appreciated🙏

0

35

95

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

📢 Dear researchers: as you know, currently our 🇺🇦 colleagues are literally being bombed 👇. @Sci_for_Ukraine is a scholar-led initiative to host 🇺🇦 refugee students & researchers. Happy to announce that I'll coordinate for 🇩🇰 Denmark. How to help: /1

@leonidragozin

Leonid ХВ Ragozin

2 years

One of the buildings of Kharkiv university on fire, hit by Russian missile. One of the first universities in Russian empire, officially founded in 1805. I interviewed a pro-Russian academic here in 2014. Putin is now destroying “Russian world” he is claiming to protect.

393

7K

19K

1

54

96

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

5 months

Do you have an idea for PhD/postdoc in NLP, and are looking for funding? Consider Denmark: @DataScienceDK call is out (deadline Feb 29) - - If you'd like to work with me, have a look here and reach out:

2

38

96

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

Yoshua Bengio, #AAAI2020 @RealAAAI : Current DL can be compared to Kahneman's System 1: fast, but heuristic-driven. We need to figure out how to get to System 2, planning and reasoning in complex situations. Comment: if you haven't read "Thinking fast and slow", you really must.

Tweet media one

Tweet media two

Tweet media three

1

24

95

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

On search engines in information warfare: this is the image search for query "Буча" (Boucha, Bucha) in @Google vs @yandex . The latter by default shows old images, and to get the recent ones you have to find the toggle in the top bar. /1

Tweet media one

Tweet media two

2

27

93

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

Doing meta-reviews for - and honestly it feels like a luxury hype detox spa. So many clear research questions and reality-check findings! I'm really proud to organize this. Save the day to learn from all the failures at @emnlpmeeting !

0

5

95

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

Has anyone noticed that Google Scholar misses citations when the official acl-anthology bibtex is used? Just found such a case. GS is aware of several abbreviated citations for a NAACL 2019 paper, but ignores a proper citation. #NLProc

Tweet media one

18

11

94

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

I am truly terrified to think that half of @emnlpmeeting submissions might be of "ChatGPT can/can't do Y" type :(

@artetxem

Mikel Artetxe

1 year

I am fed up with all those "how good is ChatGPT at X" papers. Sharing some fresh ideas: - Evaluating Siri's ASR in noisy environments - How accurate is background removal in Zoom? - Google Translate in the biomedical domain - Can Gmail detect LLM-generated spam?

9

28

384

6

6

92

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

In festive🎄angst news: SuperGLUE is officially "solved". Twice. Not sure if 0.7% improvement over T5 means NLP got radically better. But it IS a mental goalpost: we're now back to defining "NLU" & how to test it 😱 A biased summary of my fav benchmark ideas from 2020👇: /1

3

24

93

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

#NLProc #metatweet : here are all the shared tasks in @EMNLP2020 workshops and satellite conferences. Each tweet links to the announcement tweet from the organizers - so any questions could be taken directly to them. There's still time to have some NLP fun this summer!

2

35

91

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

The face of ML reality.

@IntuitMachine

Carlos E. Perez

3 years

Tweet media one

1

12

101

3

10

92

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

#NLProc Twitter suggestion: Self-promotion paper tweets look (and feel) like shameless plugs, but are crucial for careers of junior & minority researchers. How about we use a special tag for paper summaries done by the authors? Say, #NLPaperAlert ? /1

9

13

91

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

5 months

Good news + bad news: next #EMNLP is in Miami. Heads up for those of us who will struggle to get visas. My own application this year took over 9 months.

Tweet media one

2

10

88

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

The preliminary program for Insights from Negative Results is up at ! If you attend @emnlp2020 , on Nov 19 you'll have a chance to detox in a hype-free, reality-check zone. On the menu: - invited talks by @radamihalcea and @byron_c_wallace with live QA /1

3

20

86

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

An objection👆: "The US/NATO has geopolitical interests and must have..." (a) hard version: "directly orchestrated all the regime changes to democracies" (b) soft version: "exercised some 'soft' influence contributing to regime changes, eg through cultural programs" /2

1

0

78

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

@aclmeeting author response has started, you should be able to see the reviews. Notifications are coming out. Two new things: 🆕recommendation score is split into 'soundness' and 'excitement' 🆕authors can flag serious review issues by type! /1

Tweet media one

11

15

79

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

@mmitchell_ai Seems like in OpenAI discourse AGI is imagined as smth "emergent". We just keep doing what we're doing, maybe at some point it'll appear. Imagined likelihood high. In contrast, documenting terabytes of text is real work, and the ease of that is estimated realistically 😂

1

4

76

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

A highlight from @zacharylipton 's talk at on positive/negative results in engineering papers: If X works, even for no reason, community wants to know about it. But not if X doesn't work - since there was no reason why it would work in the first place. /1

Tweet media one

1

9

79

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

New blog post: my field notes from #EMNLP2021 on things that worked and didn't work in our first hybrid conference, from the on-site attendee perspective.

Tweet card media

Field Notes on Hybrid Conferences (EMNLP 2021)

Field notes from EMNLP 2021, the first hybrid *ACL conference.

hackingsemantics.xyz

6

6

79

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

10 months

@mar_kar_ @boydgraber @chokkanorg & yours truly wrote a 36-page #ACL2023 peer review report! Summary: Full report: We're pleased to clear the good name of #Reviewer2 . The authors actually complained the most about #Reviewer1 ! /1

Tweet media one

5

18

78

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

I'm considering a post on reviewing resource papers at main *ACL conferences. Dear #NLProc data people, any invalid criticisms you've been getting often? I'll start: * "too many datasets already"; * "datasets should go to LREC/workshops"; * "the new dataset is not the biggest".

10

12

77

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

Why, oh why do we limit the number of people we can thank? It's not like 2020 is too rich in positive emotions... #EMNLP2020 #CameraReady

Tweet media one

4

5

76

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

#NLPaperAlert : QA Dataset Explosion🔥is out in ACM CSUR, updated with 30+ new resources! with @nlpmattg @IAugenstein We surveyed 200+ QA/RC datasets to develop a taxonomy of formats & reasoning skills. Also discussed: modalities, domains, non-English data

Tweet media one

2

23

75

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

#coling2020 Should #GPT3 have the right to Free Speech? Panel with Robert Dale, Emily M. Bender ( @emilymbender ), Pascale Fung ( @pascalefung ), Christopher Potts ( @ChrisGPotts ) Best attempt at summary, corrections welcome. /1

3

19

72

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 month

📢 Looking for a PhD student to start in Sep 2024! Applications due April 15 This project is about LLM competence in interpreting prompts and responding to them: If you're at @eaclmeeting , you can catch @robvanderg who will co-supervise this project!

PhD Student: Linguistic competence of language models in prompt interpretation and text generation

candidate.hr-manager.net

0

27

75

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

After pulling some hair out, and before plunging into #EMNLP2021 rebuttal: Remember that their review policy DOES link to the EMNLP2020 bias list, including the all-time #NLProc favorites like "not SOTA", "too simple", "too narrow", "not DL", & "data work should go to LREC". /1

@firojalam04

Firoj Alam

3 years

I want to keep twitting this article till our #NLProc community get used to not reject a paper arguing something like " @EMNLP is not a suitable venue for a paper whose main contribution is a data set" what do you say @annargrs ?

0

5

32

6

18

72

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

Some highlights from keynote by @tallinzen at : Instead of learning the target task, models might learn the process that generated the particular dataset (for crowdsourced data it's often quite simple). Diagnostic: test sets for specific capabilities. /1

Tweet media one

2

10

72

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

When BERT Plays the Lottery, All Tickets Are Winning! New preprint with @sai_prasanna @arumshisky TLDR: based on importance scores, there are "good" and "bad" subnetworks in fine-tuned BERT's heads and MLPs. But the "bad" ones do well if re-trained!

Tweet media one

3

9

67

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

Good news 🥳: will return for the 3rd iteration, this time co-located with @aclmeeting in Dublin! Teaser: invited insights by @tallinzen @barbara_plank ! This gives us about 6 months between 2nd & 3rd workshops. Time to save up any interesting failures😉 /1

1

10

72

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

Insights from Negative Results will be back at @emnlpmeeting ! As last year, the workshop welcomes both archival and non-archival short papers. The goal is to not just provide a venue for bad news, but to organize meaningful reality-check discussions. /1

Workshop on Insights from Negative Results in NLP

insights-workshop.github.io

2

15

70

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

Ok this is real. One confirmed sighting of #Reviewer2 demanding ChatGPT already at #ACL2023 ! In which case, it's not only scientifically dubious 👇, but also breaks the 3-month "contemporaneous work" rule. ⚠️ @emnlpmeeting @ARRPreprints , this #Reviewer2 is heading your way...

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

1 year

Will reviewers now ask "where's GPT4?" We propose a #BenderRule -like rule: That which is not open & reproducible* can't be a requisite baseline w @b_niranjan @LeonDerczynski @JesseDodge @alkoller @SashaMTL @MaartenSap @royschwartzNLP @nlpnoah @strubell

10

101

370

4

4

69

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 months

🏆 I am a #VillumYoungInvestigator 2024! Tons of gratitude to @VELUXFONDEN @VILLUMscience for supporting my moonshot towards more transparent generative AI systems. I'll be looking for two PhD students & postdoc soon to join @ITUkbh @NLPnorth , stay tuned!

@VELUXFONDEN

VELUX FONDEN

3 months

Meet our #VillumYoungInvestigators 2024 - the research stars of the future. We have granted a total of DKK 132 million to 18 top talented researchers. Their #science projects range from new PtX-systems to more transparent chatbot systems. #dkforsk #dkvid

Tweet media one

1

10

52

4

5

69

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

Dear #emnlp2020 workshop organizers: let's talk about Findings. Initially only a few papers were supposed to get there. But the acceptance numbers are: main track - 754, Findings - 520, rejected - 1840. That's a lot of good papers to not get presented - more than ACL 2018! /1

6

12

69

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

Attending @emnlp2020 ? Save the date (Nov 19) for some negative results! - invited talks by @radamihalcea and @byron_c_wallace - insightful failures in QA, NLI, dialogue... - panel on leaderboardism with @ethayarajh , @JesseDodge and @rctatman

2

17

68

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

🥰🤗💐

@emnlp2020

emnlp2020

4 years

[UPDATE] The camera-ready instructions were modified. Acknowledgements NO LONGER COUNT towards the total page limit! So feel free to thank as many people as you like :)

Tweet media one

5

20

167

0

5

69

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

Happy to be part of "Rethinking ML papers" ICLR workshop team! If you've been sharing research in a format other than paper pdf - we want to hear about your experience! Anything goes: interactive pages, visualizations, audio/video, comics... Existing projects welcome as exhibits.

@rethinkmlpapers

Rethinking ML Papers

@rethinkmlpapers

3 years

Our call for exhibits, workflows, and position papers is open. Share with us your thoughts on the future of scientific publishing formats. Submissions can be webpages, blogs, notebooks, comics, podcasts, and much more! Retweets appreciated (1 / 6)

Tweet media one

2

39

120

1

15

68

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

If you've missed this gem - you really should see it. This is what 1935 must have felt like.

@just_whatever

Michael Elgort

2 years

I have translated and added subtitles to the latest video speech by Vladimir Putin from two hours ago. Please don’t let it go in vain - I want everyone to see what a speech of true fascism looks like. No further comment needed, it’s all here, in his speech

5K

23K

50K

3

13

66

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

3 years

BERTology might actually become a science! Fingers crossed for a stream of studies replicating the findings for the original model instance.

@ThiboIbo

Thibault Sellam

3 years

BERT's performance varies when you re-train it. How much does this affect your results? We release MultiBERTs: 25 pre-trained checkpoints and a statistical library to support robust research on BERT. Paper: Models: #NLProc 1/X

Tweet media one

6

77

365

0

8

65

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

A great opinion piece on #leaderboardism in #NLProc by @ethayarajh and @jurafsky : Title: Utility is in the Eye of the User: A Critique of NLP Leaderboards Preprint: /1

Tweet card media

Utility is in the Eye of the User: A Critique of NLP Leaderboards

Benchmarks such as GLUE have helped drive advances in NLP by incentivizing the creation of more accurate models. While this leaderboard paradigm has been remarkably successful, a historical focus...

2

14

64

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

For the #EMNLP2022 authors new to the field: here's an evergreen guide for responding to reviewer comments Note that softconf imposes a limit on the total number of words in the response.

Tweet card media

How we write rebuttals

By Devi Parikh, Dhruv Batra, Stefan Lee

deviparikh.medium.com

0

10

64

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

2 years

🏅 @meta stopped monetizing or selling ads to Russian state media: /2

@ngleicher

Nathaniel Gleicher @[email protected]

2 years

1/ We are now prohibiting Russian state media from running ads or monetizing on our platform anywhere in the world. We also continue to apply labels to additional Russian state media. These changes have already begun rolling out and will continue into the weekend.

951

2K

12K

1

2

64

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

In other news: GPT-3 grand debut on superGLUE didn't go too well. It's currently the worst non-baseline system, although it does pretty well on COPA and ReCoRD.

Tweet media one

7

5

62

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

I do hope that the ACL peer review reform also includes a move to something better than softconf.

@emnlp2020

emnlp2020

4 years

Dear all, we're hard at work to send out the notifications within a couple of hours. However, softconf is very slow and unresponsive - presumably due to all of you trying to log in to check the results. Please don't, and the emails will be out soon.

2

26

143

1

4

61

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

Your daily reminder that syntax has meaning! #ConstructionGrammar

Tweet media one

2

2

59

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

Come work with me! Copenhagen is awesome, @CPH_SODAS has experts on every aspect of human existence, plus collaborations with @coastalcph and @CopeNLU . Oh, and Denmark actually processes visa applications - I got mine in NY in May!

@daviddlassen

David Dreyer Lassen

4 years

Want to help integrate data science into social science? We are hiring 1-2 postdocs for the UCPH Social Science Data Lab at @CPH_SODAS and @KuSamf under the U. of Copenhagen Data Lab initiative. Deadline Aug 1, details here: Please RT! Email/DM if Qs

1

62

52

0

14

60

@annargrs

Anna Rogers 🇺🇦🇪🇺 is looking for postdocs!

4 years

To all the happy authors of the papers accepted to Findings of EMNLP: the workshop for Insights from Negative Results will reopen submission for non-archival abstracts. Presentation&discussion shouldn't be a priviledge. /1

Workshop on Insights from Negative Results in NLP

insights-workshop.github.io

1

8

59