We've spent a tremendous amount of time reflecting whether the NLP task of Sentiment Classification (x=review, y=rating) is causal or anticausal since 2020. Check out our 2024 latest answer➡️
💡We combined Causality and Psychology insights & improved
#LLMs
!
Many students aiming to apply for PhDs devote this summer to full-time research. I just took a chance to re-organize our repo of many "how to do research" suggestions: You can learn from great minds 🧠& stand on the shoulder of giants💡!
#NLProc
#AI
There’s no paper deadline on Christmas Eve. But yesterday was both the
#ACL2023
deadline and the Chinese New Year’s Eve, the biggest get-together-with-family event for many ethnically Chinese people. I know organizers are doing hard work, but maybe more future D&I can be great🥰
Really excited that our paper "A PhD Student’s🧑🎓 Perspective on Research in NLP in the Era of
#LLMs
" will be at
#COLING2024
! We brainstormed 45 topics💡 that students can work on despite the LLMs. Co-led by
@OanaIgnatRo
@ZhijingJin
and
@radamihalcea
with contributions from many.
“What should I work on?” is a question we hear more & more often from NLP students, during a time when the media rhetoric is that “it’s been all solved”
Turns out there are many NLP research areas rich for exploration—here is our answer from 20+ students
Our
#NeurIPS2022
oral paper asks a question:
- Can we say we’ve defined our tasks appropriately for
#AI
/
#NLProc
?
- The answer is "No." If a model executes "literally" what we told it to do, there might be serious risks.
- [Vision] We need to teach our models to handle exceptions!
Waking up this morning, I saw the first email in my mailbox the "EECS Rising Stars🌟2023 Acceptance" for my research "Socially Responsible NLP via Causal and Moral Principles🧠". Truly humbled for this recognition, and thankful to all who've supported me❤️on my research journey!
I keep thinking about how to help junior
#NLProc
and
#AI
researchers (with our initiative
@aclmentorship
-for-all). One's career path largely depends on "information spread" which is unequal due to the social network we are in. Repo for equal access to info:
I know I'm being emotional. But the neighborhood area of my family in Shanghai has been locked down for several days. My parents are not allowed to go buy food. They are running out of vegetables and will probably eat our storage rice. I want to take the courage and say: /1
Arriving at Seattle today! I will start my internship with Mona Diab
@MetaAI
from now until Sept 17. I will work on
#CausalNLP
for the Responsible AI Team. Happy to connect with people in the Seattle-Bellevue area!🎉
Unsatisfied with citations as a measure of paper quality? Check out CausalCite, an alternative metric where we quantify paper impact by its causal effect on followup works. We ask the counterfactual question, "Had this paper not existed, what would happen to the followup papers?"
Excited to announce that I'm on the
#NLProc
job market this year! Actively seeking assistant professorship positions with my research on Causal NLP and Responsible AI. Looking forward to new opportunities and collaborations! ➡️
#JobSearch
#AcademicTwitter
Can LLMs pass 𝐂𝐚𝐮𝐬𝐚𝐥 𝐓𝐮𝐫𝐢𝐧𝐠 𝐓𝐞𝐬𝐭? In our
#NeurIPS2023
paper “CLadder: A Benchmark to Assess Causal Reasoning Capabilities of LMs”, we test LLMs on the Pearl's 3 Rungs of Causality, and found that LLMs are *far* from solving them.➡️
In this era of rising political challenges and international conflicts, I highly prioritize "NLP for Policies" as an important contribution to NLP for Social Good
@nlp4sg
. So glad to author the
#NLProc
chapter in the "CSS for Policy" book w/
@radamihalcea
I’m strongly against authoritarianism. Also, many people here in Europe asked why I work a lot. I think the answer is, we never know when the freedom might end. Even the free air and free academia are luxurious to me. /2
#NLProc
has been making great advances, but we're still to learn its social implications.
@radamihalcea
and I are collecting ideas to help define the emerging concept "NLP for Social Good": -we'll anon.compile all answers and share back. RTs appreciated❤️!
Finally finished the EMNLP-NeurIPS-NAACL ddl bundle! So proud of our wonderful org. team w/
@YejinChoinka
et al. @ Moral AI (MP2) Workshop; our CLadder paper team; our NLP4SG initiative+paper; our
@ACLMentorship
BoF; our
@BlackboxNLP
panel;&lovely memories w/ you all in SG&NOLA!
So let me still do my best, in the best years of my twenties, to devote this free self to my beloved research. And let life take over for all the uncertainties ahead. At least, no regret for these wonderful years of youth. /end
For people interested in
#CausalNLP
and Interpretability & Robustness of
#LLMs
, I just shared my Keynote Talk at
#EMNLP2023
@BlackboxNLP
workshop: "Causal NLP: A Path towards Opening the Black Box of NLP"
- 🎦Talk video:
- 💻Slides:
The Dream-Comes-True❤️ moment for causality researchers -- meeting Judea Pearl
@YudaPearl
in person! Yuda's profoundly knowledgeable in philosophy, genuinely enjoys singing, and is just charming. Truly touched by his talk & personal stories w/
@kahneman_daniel
(Two great minds!)
It's such an honor to supervise my talented Master student
@flavioschneide
at
@ETH_en
@ETH_AI_Center
. See our "audio-diffusion-pytorch" Python package (🔥167 stars now🌟) -- 1st Diffusion Model for Audio (music/speech/...) 🥁🎵🎻🎙️ that works well! Stay tuned for 3 use cases:🧵=3
We are organizing a special edition of ACL Mentorship
@emnlpmeeting
#EMNLP2023
in Singapore:
"Sharing Learnings from This Conference & Identifying Promising Research Directions" w/
@radamihalcea
@mohitban47
@yuntiandeng
! Welcome all to join us Dec 9 (Sat) 1:45pm in Aquarius 1🎉!
Should reviewers🧑⚖️ know each other's identities during the discussion phase? We conducted an RCT in 𝗨𝗔𝗜 𝟮𝟬𝟮𝟮 investigating the causal effect🪄 of anonymity in reviewer discussions on reviewing outcomes.💡Key findings: (1) Anonymity leads to slightly more discussions. 1/4🧵
Excited for our paper "Beyond Good Intentions: Reporting the Research Landscape of NLP for Social Good (NLP4SG)" at
#EMNLP2023
Findings. We propose 3 NLP tasks: (1) detect if a paper relates to NLP4SG , (2) check which UN SDG it addresses, and (3) extract task & methods proposed.
Experiencing the "survival crisis" as a PhD student in
#NLProc
due to LLMs? Together with
@radamihalcea
and 20+ friends at the LIT lab of
@michigan_AI
, we provide a list of 48 NLP research topics across 15 areas you can do📜! Lots of hope around🌟!
“What should I work on?” is a question we hear more & more often from NLP students, during a time when the media rhetoric is that “it’s been all solved”
Turns out there are many NLP research areas rich for exploration—here is our answer from 20+ students
Excited for EMNLP and NeurIPS…And, I’m on the job market–welcome any chat! Here’s what’s coming at
#EMNLP2023
:
➡️Welcome to my Keynote on Causal NLP
@BlackboxNLP
at 9am Thu & Panel
➡️Our poster on NLP4SocialGood 10 Dec 11am
➡️Our
@ACLMentorship
BoF “Sharing Learnings from EMNLP”
☀️Good news comes in pairs: Our (1) NLP4PositiveImpact workshop
@NLP4PosImpact
@nlp4sg
and (2) CausalNLP tutorial
@CausalNLP
are both accepted at EMNLP2022
@emnlpmeeting
. Thx to my wonderful co-organizers to make this happen!💕🤗Also a big thank to the EMNLP organizers!
Extremely honored to be selected to attend the prestigious 73rd Lindau Nobel Laureate Meeting
@lindaunobel
as a Young Scientist. Looking forward to this incredible once-a-lifetime event connecting with 30 Nobel Laureates🌟 for a week!
#LINO24
#WomenInSTEM
Today, we invite more than 650 Young Scientists from 90+ countries to meet well above 30 Nobel Laureates at the 73rd Lindau Nobel Laureate Meeting (30 June–5 July 2024 |
#LINO24
|
#Physics
). We can‘t wait to welcome you all to Lindau!
I know many of our friends&family are suffering from this tech decline. I’m so sorry. While thinking about potential solutions, maybe seriously consider AI+X, where X is a domain you have faith in. And always reason about our society in 10 years, 50 years, and >100 years.
Little tip: Google Scholar profiles should link with your personal Gmail, but not changeable org emails. I just updated my profile to a new link . When updating it, I can't help being emotionally touched & thankful to all who helped my growth in research🤗!
Humbled to receive 3 Rising Star🌟 awards this Fall. I look forward to the upcoming 2 weeks at Georgia Tech (this week), U Chicago (early next week), and UMD (mid next week) for their mentoring workshops. Really thankful for this opportunity & happy to chat with people around!
We're proud to announce the winners of our Rising Stars in Machine Learning program, designed to support researchers from underrepresented groups.
Pictured clockwise: Megha Srivastava, Han Shao, Wanrong Zhu, Laixi Shi, Zhijing Jin, Yutong Bai, Weijia Shi, and Sanae Lotfi.
I am excited to continue my work on NLP for social good
@nlp4sg
and
@CausalNLP
with the generous support from 2 PhD Fellowships: Many thanks to
@open_phil
Open Phil AI Fellowship & Future of Life Institute PhD Fellowship granted by
@tegmark
@FLIxrisk
! 🎉
Please welcome our first class of
@VitalikButerin
Fellows in AI Existential Safety, and consider collaborating with them to ensure that AI become the best rather than worst thing to happen to humanity!
Humbled to receive the "Responsible AI Award" at the Michigan AI Symposium
@michigan_AI
. Grateful for this amazing venue for Responsible AI discussions & idea exchange. Together w/ our
@NLP4PosImpact
Workshop efforts, happy to foster more joint efforts on NLP for Social Good🥰!
Thinking about being a prof in the future, I also wonder "At the end of the year, what'd be good to say to the labmates?" A good idea is to share lots of positive feedback, but I'm also touched by a profound email with care for humanities and the mission of scientists.
I'm co-teaching an
@ETH_en
seminar course on
#GenerativeAI
today & tom, 9:30am-6pm in Zürich. You're welcome to email me if you need Zoom access to listen to our seminar "Representations in Generative AI: Causal Methods, Images, Music&Language"! Welcome all interested students!🎉
🙌Last call for our survey to define “NLP for Social Good”: ➡️Please voice your opinion!🤗Your response will contribute to the NLP for Social Good discussions at the upcoming ACL 2021
@aclmeeting
&
@NLP4PosImpact
!❤️--Warm thanks by
@radamihalcea
@ZhijingJin
[Call for Pull Requests] I've been curating a list of NLP for Social Good
@nlp4sg
papers , so that interested people can just save their time. Soon I found this task gets too heavy, so I want to call for
#NLProc
community to PR&add NLP4SG papers you enjoy!
Truly astounded and angry about the Hamas terrorism attack on Israel. Extremely brutal and inhumane. For all the Israeli members in our community: if you need to shift your workload, feel free to let me know if I can help with anything. Any organizational or paper-related stuff.
2 ways to test LLMs: (1) test it on many datasets, (2) check what's happening inside. For (2), I'm very impressed by "mechanistic interpretability." I explored 2 behaviors below, and find the "skip trigram" in Transformer Circuits by
@ch402
very convincing to explain them 🧵
End of the year is a good time for more thinking and reflecting. -- I'm freeing up more time for non-paper-driven research brainstorming. Happy to chat about (or if I can help, then advise on) any of the following (about deeper science,
#Causality
,
#NLProc
#AI
, and society): 🧵/5
Happy to share some tips after a series of invited talks in Ann Arbor-Boston-Saarbrücken. Job search can be exhausting, but apart from managing talks and 1:1s, don't forget: (1) take good care of yourself❤️(2) be kind & grateful to hosts🤗(3) travel w/ a light heart and luggage🍀
Can LLMs pass 𝐂𝐚𝐮𝐬𝐚𝐥 𝐓𝐮𝐫𝐢𝐧𝐠 𝐓𝐞𝐬𝐭? In our
#NeurIPS2023
paper “CLadder: A Benchmark to Assess Causal Reasoning Capabilities of LMs”, we test LLMs on the Pearl's 3 Rungs of Causality, and found that LLMs are *far* from solving them.➡️
Thx to the active feedback for our 𝗖𝗼𝗿𝗿𝟮𝗖𝗮𝘂𝘀𝗲 paper "Can LLMs Infer Causation from Correlation?" (), we just released v2.0 with 5 more LLMs (e.g., LLaMa2), now totaling 17 off-the-shelf + 17 finetuned LLMs, and also removed duplicates in the data.
I know many students are stressed by the ongoing grad school application season. We will organize an ACL Mentorship session on Dec 1 on "Applying for a PhD in NLP." To promote global equality, we welcome mentors & mentees from all over the world.
[Upcoming] Applying for a PhD now? On Dec 1, we will have a mentoring session on "Applying for a PhD in NLP." Our 3 mentors on the 1-hour panel are Prof
@radamihalcea
(UMich), Prof
@mrinmayasachan
(ETH), Prof
@ikekong
(HKU). Submit your Questions to slido
Can LLMs pass the Turing test for causality? Or what should be the ultimate check for its causal reasoning ability? We just had a nice talk series on “LLMs+Causality” at the Causal Data Science Meeting 2023. FYI, here are the papers for the 4 talks:
Applying for grad schools this year? We are organizing a special
@aclmentorship
session on grad school applications in
#NLProc
next Tuesday w/ Prof Nathan Schneider
@complingy
, Prof
@LukeZettlemoyer
, and Prof
@yuntiandeng
. Welcome to join the live session to ask anything! 🧑🎓
📢 Join us for the next ACL Mentorship session: "Preparing your grad school application: CV and Statement of Purpose"
📷 Tu October 10th, 6:00-7:00pm UTC / 2:00-3:00pm EST / 7:00-8:00pm GMT
Zoom:
At the upcoming
@NeurIPSConf
, I’m excited to chat w/ departments interested in hiring
#NLProc
and
#ResponsibleAI
. Also,
✅Check out our Poster “CLadder: Causal Reasoning in LLMs” Dec 12 Tue 10:45-12:45
✅Welcome to our Workshop on Morality & AI Dec 15 Fri
For people who care about
#NLP4SocialGood
#ResponsibleAI
and related topics, we are glad to announce that our next
@NLP4PosImpact
Workshop is accepted at
#EMNLP2024
! Feel free to submit your paper to ARR, and commit to our workshop when the CfP is out. See you next year in Dec!🥰
I'm presenting our
#EMNLP2023
poster "Beyond Good Intentions: Reporting the Research Landscape of NLP for Social Good" at 11AM - 12:30PM today! If you are interested in the current progress of NLP for Social Good research landscape, welcome to drop by!
Excited for our paper "Beyond Good Intentions: Reporting the Research Landscape of NLP for Social Good (NLP4SG)" at
#EMNLP2023
Findings. We propose 3 NLP tasks: (1) detect if a paper relates to NLP4SG , (2) check which UN SDG it addresses, and (3) extract task & methods proposed.
[
#acl2021nlp
Note Series] Excited about
@radamihalcea
's ACL Presidential Address
@aclmeeting
! Don't just chase "accuracy," but care about "more dimensions": interpretability, generalizability, ethics, NLP for social good, environmental cost of NLP, D&I, etc! /1
Key info for people coming to
#EMNLP2022
: (1) Abu Dhabi is very safe. OK to walk or Uber at late night. (2) Hair cover is not mandatory for foreigners. (3) Get sun scream for the hot sun. (4) Aloft is "inside" the Exhibition Center, 2nd floor. (5) Covid test is provided by EMNLP.
Right on time for NAACL -- We just compiled our 2 new episodes of ACL Year-Round Mentorship sessions: "Finding a Great Advisor" (with suggestions about reaching out to advisors) and "Surviving Grad Schools" (about taking care of yourself during research):👉
Bernhard's talk always inspires me with profound knowledge connecting
#CausalML
w/ physics and philosophy. Causal models lead to better out-of-distribution generalization, modularity, sample-efficient learning, and explainability.
@bschoelkopf
@Conf_CLeaR
2 more tips:
- Broaden our knowledge. Don't just study CS, but PPE: philosophy, politics and economics. Care deeply about our society.
- Think realistically about our hierarchy of needs. We're first a human, a family, a friend, a part of our society, and then a CS person.❤️🌱
Another v2.0: We just largely improved our CausalCite's writing! New plots:(1)Our research question is "had this paper never been published, what would have happened to its follow-up studies?"(2)We use graphical criteria to control for the confounding vars
Just arrived in the Netherlands for
#UAI2022
. Excited to talk to & learn from causality people here! 💡Keywords I'm interested in: causality, robustness, epistemology, language models. I'll also have a poster at the Causal Representation Learning workshop
@crl_uai
. See you there!
Thank you all who participated in our BoF on
@nlp4sg
today at
#ACL2021NLP
@aclmeeting
!! So many ideas, suggestions, and ways forward from 50++ participants! We will send the compiled notes to those who are interested. Let's keep up the momentum and use NLP in an meaningful way!🙌
Hearty congrats to my dear mentor
@radamihalcea
for being appointed as Janice M. Jenkins Collegiate Professor, and her wonderful talk on “Natural Language Processing with People, for People, by People.” I really admire her support for Women in CS and NLP for Social Good
@nlp4sg
!
Yesterday was a special day: I got to meet my role model Prof. Janice Jenkins — the very first woman faculty in EECS at Michigan, a pioneer in CS & biomed and an active supporter of women. Janice’s story is inspiring:
I also got … a chair 🪑😊
Childhood dream one step closer - impacting society with our ideas & speaking to policymakers. Honored to present my
#CausalNLP
research
@MPI_IS
to State Minister-President
#Kretschmann
, Research Minister&State Secretary. Grateful for each milestone!🙌
"What do people think of when we talk about NLP4SocialGood?" We've been working on this since 2021, our 1st
@NLP4PosImpact
Workshop. Now we proudly introduce where we visualize all
#ACLAnthology
papers and what kind of Social Good goal they target at❤️!
What impact does NLP have on Social Good (NLP4SG)? In our new paper, we introduce a new scientific NLP task and a data visualization that shows an overview of the NLP4SG research landscape. (See our live demo: 💙)🧵
#NLProc
Our mentorship team
@aclmentorship
is here at 10ET tomorrow at
#ACL2023NLP
! Happy to help everyone seeking career advice, having questions, or eager for a personal connection w/ the community 🥳. Also welcome senior members to drop by (Dockside 2) and be advice-givers!🥰🪩
Meet our "ACL Mentorship" community
@aclmentorship
at the BoF Meetup on Wed 10ET
#ACL2023NLP
@aclmeeting
!
🎯Target Audience: Everyone seeking career advice, having questions for senior researchers in
#NLProc
, or eager for a personal connection with the community 🥳
Great to have published this work at Natural Machine Intelligence jointly w/ an awesome team led by
@_dieuwke_
at Meta AI
@AIatMeta
! We create a taxonomy for out-of-domain generalization for
#NLProc
: . Also welcome to our GenBench Workshop at
#EMNLP2023
🎉!
Published today in Nature Machine Intelligence — GenBench is an effort led by AI researchers at Meta that aims to make state-of-the-art generalization testing the new status quo for NLP work.
Read more about the work in
@NatMachIntell
➡️
Thanks a lot to
@lucie_nlp
for inviting me to talk about
#CausalNLP
at Marburg University
@uni_MR
in Germany. The research discussions with you and
@CaisaLab
also gave me much inspirations for this
#CausalNLP
research path forward!
[ACL-IJCNLP 2021 Note Series]
@aclmeeting
is starting now! Interested in what are hot topics in NLP? The chairs just introduced a trending keyword list: BERT, named entity recognition (NER), language models, machine translation, attention weights, ..., Covid-19 research, etc! /1
🚀 Our newest video "Doing NLP on a Budget" featuring
@eltimster
@LucianaBenotti
@vinodkpg
is out!
- What are good research topics w/o tons of GPUs?
- Should I run all the human evaluation asked by reviewers?
- Will NLP be more & more annotation-intensive?
I want to say a sincere Thank-You to Prof
@drvmago
for inviting me for a 1.5-hour talk on NLP for Social Good
@nlp4sg
at Lakehead University
@mylakehead
(Canada), with an amazing audience of 150 researchers from different Uni's.🤗An open-to-all seminar:
Feeling the language barrier of AI research? NLP is here to help! Join
@MonaDiab77
@bkhmsi
& others at the "ACL 60-60 Initiative"
@naaclmeeting
#NAACL2022
tomorrow. To democratize research to non-English communities, ACL 60-60 translates research to 60 languages+sign language 1/2
Join us at
#NAACL2022
, Wednesday July 13 from 3-4PM PT, for a hybrid (physically in room 506 Samish, or virtually on Underline) Birds of a Feather discussion on the future of the ACL 60-60 Challenge/Initiative. Spread the word! Come one come all!
😊The Virtual part of our NLP for Positive Impact Workshop
@NLP4PosImpact
at
#EMNLP2022
@emnlpmeeting
is happening now!
- Links: See our calendar events!
- Full workshop schedule:
I just listened to and was really impressed by
@lilyxu0
's PhD work on AI for Social Good! Amazed at her extensive work into various aspects of human health and planetary health. And it's great to see PhD research combined with *field work* to regions in need too! Love this!🥰
I'm on the academic job market!
I'm a PhD student at
@Harvard
, where I develop and deploy AI decision-making techniques for planetary health.
My research enables practitioners to take efficient, robust actions necessary in these high-stakes, low-data settings.
We are organizing a special edition of ACL Mentorship
@emnlpmeeting
#EMNLP2023
in Singapore:
"Sharing Learnings from This Conference & Identifying Promising Research Directions" w/
@radamihalcea
@mohitban47
@yuntiandeng
! Welcome all to join us Dec 9 (Sat) 1:45pm in Aquarius 1🎉!
🔊 Happening in 3 hours 🔈
Join today’s ACL mentorship session at 11am EST / 4pm UTC to learn more about the PhD graduate job market. Equally relevant for those about to graduate with a PhD or those considering whether to enroll in a PhD program.
My labmate
@NJingwei
did a wonderful job exploring "When Does Multi-Task Learning Work?" with extensive experiments across Financial NLP tasks. We look into factors such as skill diversity & task relatedness, and share our insights in the
#ACL2023
paper 📈
Thank you for the great invite! I’m excited to talk about my CausalNLP research in one of the rising star talks at
@michigan_AI
Symposium on 17th (Tue). See you in Ann Arbor! (I’m also happy to take chats throughout the upcoming week at UMich!)
Here is our ACL Mentorship session on Grad School Applications. Good luck to all students! I had lots of ups and downs when applying for my PhD. Looking back, I’m really thankful to all who helped me, and now we wish to pass these mentorship suggestions on to more students 🥰!