Srijan Kumar - Lighthouz AI (YC S24) 🤖 @srijankedia Twitter profile

Last Seen Profiles

@wackykiray

@Lady_ice_Dinou

@Winter_nt99

@cukienaknikmati

@MickHeroNG

@FPVaughanIII

@browningbae

@furinkan

@konyk_1018

@c2arobert

@RabbitMQ

@kana_859

@sexmaster42069

@NotNot4me

@EJupitasari

@Dr_AlTannak

@Botoliya_1996

@YaduvansiAnant

@cloudy96_

@panicusersu

@quartazo

@aokisayaka0327

@XiaoCha520

@BrBomboclaat

@anon_codex

@CoachWesGBurg

@jaggupk41835

@flwrminz

@PacoBanos_MX

@boomshootart

@galery_basah10

@kichi356

@SkateCanada_NL

@mj_baile

@yomama0920

@Perth_Traffic

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

5 years

I am thrilled to announce that I will join Georgia Tech as a tenure-track Assistant Professor from Jan 2020! My lab will focus on machine learning, data science, and cybersecurity. Looking forward to working with amazing students and colleagues! @gtcomputing @GTCSE @mlatgt

48

8

400

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

Please RT! I am looking for PhD students to join my research group at @GeorgiaTech . My group works in cutting-edge topics in data mining, ML, graphs, recsys, adversarial learning. We focus on applications in social media, public health, and security. Please get in touch!

7

184

395

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

I am super excited to receive the NSF CAREER award on "Social Response-Powered Misinformation Detection, Robustness, and Correction"! 🎉

32

8

326

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

I'm extremely honored & delighted to be named in the @Forbes 30 Under 30 list in the Science category for my work in social media safety and integrity! 🎉🎉 Very thankful to @GeorgiaTech , my mentors, collaborators, students, and my family for their support in my journey!

39

14

268

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

11 months

I'm *hiring* AI engineer interns for my generative AI startup! Position is remote and for full-/part-time. My startup is revolutionizing AI product reliability and security! Evaluating LLM-based AI products and securing them in production is the missing piece of the puzzle in

33

55

255

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

11 months

Multimodality is the future! Here are 10 multimodal models that can be used to build futuristic #AI apps: Open-source: 1. LLaVa - LLaVA is the most popular open-source multimodal framework. 2. OpenFlamingo: 3. Idefics-80b:

4

65

227

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Congrats to my Ph.D. student Bing He on a "slam dunk" paper acceptance at @TheWebConf -- all four reviewers gave the highest-possible score to the paper!! #thewebconf2023

7

3

209

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

9 months

LLMs are ultra-sensitive to prompt formatting, giving accuracy ranges from 4%-88% for a given task with LLaMA-2-70B and 47%-85% with GPT3.5. Latest research by researchers at UW, UCB, and AI2 quantifies how spacing, punctuation, casing, etc. impact LLM performance. Great work

2

48

203

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

Excited to win a Facebook Faculty Research Award to work on adversarial machine learning for social media integrity w/ @PoloChau ! cc/ @GTCSE @GTCSE @mlatgt

Announcing the winners of the 2020 request for proposals in applied statistics - Meta Research |...

In February 2020, Facebook launched the Statistics for Improving Insights, Models, and Decisions request for proposals (RFP). This RFP was...

research.facebook.com

17

10

188

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

Thrilled to get my first NSF grant! We will study the impact of the COVID-19 pandemic on psychological well-being, anxiety, and distress: . Excited to be a part of a truly multidisciplinary team: @munmun10 @dvshah , Patricia Cavazos-Rehg, and Sijia Wang!

9

3

159

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

11 months

Super excited🤩to start as the co-founder and CEO of a stealth-mode generative AI startup that is revolutionizing AI product reliability and security! Evaluating LLM-based AI products and securing them in production is the missing piece of the puzzle in the enterprise adoption of

8

5

134

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

Looking forward to giving a Keynote Talk at ** The KDD 2021 Workshop on Machine Learning in Finance ** I will talk about **Trustworthy Machine Learning for Fraud Detection** Tomorrow at 8am ET.

4

23

122

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

Excited to receive *two* Microsoft AI for Health grants to support our work in fighting COVID-19. Thank you @Microsoft @MSFTResearch @Azure !

6

2

118

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

Creating mis/disinfo images just became a whole lot easier. Thank you, #AdobeFirefly , and thanks also for promoting this (mis)use case prominently.

John Nack

@jnack

1 year

This is the single most chaotic use of #GenerativeFill I’ve seen so far. 😅 #AdobeFirefly

59

376

3K

10

30

114

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

5 years

Pls RT! I am admitting PhD students to my lab @GeorgiaTech ! We will invent the next generation solutions to create a better web and society. Topics: Data Science for trust & well-being; ML for robust user models & recsys; Adversary modeling. More details

1

64

112

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

10 months

Must watch! @karpathy 's 1 hour lecture on "Intro to Large Language Models" covers: 👉 Inference 👉Training 👉Fine-tuning 👉RLHF 👉Scaling Laws 👉Multimodality 👉Self-improvement 👉Security (Jailbreak, prompt injection, poisoning)

[1hr Talk] Intro to Large Language Models

This is a 1 hour general-audience introduction to Large Language Models: the core technical component behind systems like ChatGPT, Claude, and Bard. What the...

www.youtube.com

3

20

108

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

Excited to be selected for @MSFTResearch Accelerate Foundation Models Academic Research program to investigate @OpenAI LLM capabilities (GPT-4, ChatGPT, etc) in answering health queries across languages! Excited to continue this in collaboration with @munmun10 & @verma22gaurav !

5

6

111

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

Proud of my Ph.D. student @verma22gaurav , who has won the JP Morgan AI Ph.D. Fellowship 2023 @jpmorgan ! This comes in addition to his selection as a Snap Ph.D. Research Fellowship 2023 @Snap ! Gaurav researches robust multimodal AI models for safety, equity, and well-being.

4

2

110

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

5 years

Very excited to join Georgia Tech!

Machine Learning at Georgia Tech

@mlatgt

5 years

We are thrilled to welcome seven new faculty members to the #MLatGT family. Welcome to Tech, Diyi, Srijan, Sehoon, Debankur, Xiuwei, Chao, and Judy! 💼:

0

5

47

4

1

106

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Simple is beautiful. Simple is memorable. Simple is impactful 🤩 Reviews rejecting solely due to model simplicity send the absolutely wrong message. They force students to build unnecessarily complex solutions, even when novel yet "simple" solutions work. Please don't do this 🙏

7

9

104

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

6 months

My first PhD student defended his thesis yesterday! 🎉Congratulations to Dr. Sejoon Oh! Sejoon has done pioneering research that makes recommender systems reliable and trustworthy. Committee members: @ConstantineDovr @chaozhangcs @McAuleyLabUCSD @berkustun

7

3

104

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Super excited for a new @CDCgov grant to develop solutions for Anti-Asian and Anti-Black violence-provoking misinformation!🤩 Special as it is my first grant as a PI bringing together a team of academics ( @munmun10 @LSchwabReese ) & practitioners ( @ADL )!🎉

Researchers to Study Connection Between Online Misinformation and Real-World Violence | College of...

www.cc.gatech.edu

6

5

103

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

🚨 New position 🚨 Please RT 🚨 I am looking for a postdoc to join us at Georgia Tech to work on exciting problems in * Recommender systems * Graphs * Trustworthy AI Please send your CV to srijan @gatech .edu by March 21, 2021. #AcademicTwitter @gtcomputing @GTCSE @mlatgt

1

50

98

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

Thanks to NSF for the CAREER award and for supporting my group's research! More details below 👇

GaTech CSE

@GTCSE

1 year

Assistant Professor Srijan Kumar is a @NSF CAREER Award recipient! Check out our story about what the award means for early-career faculty and how @srijankedia will use the grant to counter social media misinformation! @gtcomputing @GTResearchNews

1

2

25

11

5

97

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

Racism is a Virus. Announcing the largest study of Anti-Asian hate and counterhate speech on Twitter during #COVID19 : 3 months 30M+ tweets 800k+ hate 200k+ counterhate tweets 87M+ node Twitter network Paper: Dataset: Findings 🧵

2

24

95

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

Can Large Language Models be used for social good? Yes! My paper shows how LLMs can counter misinformation. Paper: Code: This is particularly timely as AI pioneers (Hinton, Horvitz) raise alarms of AI misuse to spread mis/disinfo

2

17

95

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

6 years

Thrilled to be awarded the 2018 SIGKDD Doctoral Dissertation Award runners-up!! #KDD2018 @kdd_news @umdcs @umiacs Many thanks to my advisor @vssubrah and my mentors @jure and @bear5claw !!

14

3

91

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

I am delighted to be selected for the AAAI New Faculty Highlights 2023 by the @RealAAAI ! I will be in Washington DC this week to present my group's research on the Advancements in AI for Web Integrity, Equity, and Well-Being.

5

2

86

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

6 months

Excited to introduce ⚔️ Chatbot Guardrails Arena ⚔️ in collab with @huggingface Goal: Jailbreak Privacy Guardrails AI chatbots have access to sensitive information. This arena stress tests LLMs and privacy guardrails in leaking this sensitive data. Put your creative caps on!

4

22

84

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

🌟Manoj's (undegrad!) 1st paper ever and @verma22gaurav 1st paper as a PhD student at @GTCSE @gtcomputing 🎊 Congratulation 🎊 🎉"Characterizing, Detecting, and Predicting Online Ban Evasion" paper accepted at @TheWebConf #WWW2022 🎉

2

0

81

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Misinformation is one of the biggest threats to today's society. Excited to be a co-PI on this $5 million @NSF grant to help journalists fight misinfo with interactive detection and mitigation tools!🎉 @gtcomputing @GTCSE @ICatGT

UW-Madison SJMC

@uw_sjmc

2 years

A group of researchers has recently received a $5 million Phase II Convergence Accelerator cooperative agreement award from the NSF. The group plans to continue developing Course Correct, a tool to help journalists identify and combat misinformation.

0

10

45

4

3

80

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

🎉 Congratulations to my Ph.D. student @verma22gaurav for winning the prestigious 2022 Snap Ph.D. Research Fellowship! 🎉 Gaurav researches AI for online safety, equity, and well-being of users, communities, and platforms. Check out some of his notable works below 👇

2

5

78

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

📢Releasing the demo of how LLMs can be used for social good to counter misinformation! This paper just got a *best paper award nomination* (top 1% paper) from @TheWebConf ! 🎊 Demo: Paper:

Can GPT/Large Language Models be used for Social Good to Counter...

This is a demo video for the ACM Web Conference 2023 paper titled "Reinforcement Learning-based Counter-Misinformation Response Generation: A Case Study of C...

www.youtube.com

4

7

77

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

What an incredible honor to meet @GeorgiaTech President @CabreraAngel for lunch! Had exciting discussions about AI, ML, Data Science, social media, psychology, and misinformation! #georgiatech #ai #ml #datascience #tech

1

77

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

Peer reviewing is arguably one of the most crucial pillars of science. Recently, I was delighted to be recognized as an *Outstanding Program Committee Member* by the ACM WSDM 2022 conference! 😁🎊 @WSDMSocial is a top data mining CS conference!

3

0

76

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

7 years

Our work on intercommunity conflicts/raids/wars on @reddit is now live: . Paper at @TheWebConf WWW 2018 w/ @williamleif @jure @jurafsky . Insights: 1% of communities initiate 74% of all conflicts, are started by active members, & can be predicted in advance

3

24

73

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Honored to be named a Kavli Fellow by the National Academy of Sciences (NAS) two years in a row, this time for 2023! @KavliFoundation @gtcomputing @GTCSE

7

3

72

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Multiple exciting postdoc openings in my group to do cutting-edge research on data science, computational social science, social media, and online misinformation! Please send your CV to srijan @gatech .edu #KDD2022 @kdd_news

2

35

69

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

Our paper on how non-expert crowd can help curb misinformation spread is influencing the idea behind Twitter's new crowdsource counter-misinformation initiative @birdwatch ! Paper: About birdwatch: @GTCSE @gtcomputing @mlatgt

1

4

70

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

5 years

Excited to give an invited talk at @GoogleAI Research on dynamic embeddings to predict online bad behavior! I talked about our recent papers at KDD 2019: & WWW 2018: . Code & data at: &

1

8

68

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

10 months

Excited 🤩 to be at #NeurIPS2023 from Dec 13-16 to give a keynote at the "Robustness of Few-shot and Zero-shot Learning in Foundation Models" workshop! Come say hi, if you are around. Building robust and reliable #AI models, including #LLMs and multimodal models, is the need of

3

5

68

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

Congratulations to my PhD student, Sejoon Oh, for passing his quals exam! 🎉🎉 @GTCSE @mlatgt @gtcomputing

3

0

67

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

8 months

Dare to ask a healthcare question to an #LLM in Hindi or Chinese? 🎊 New paper "Better to Ask in English: Cross-Lingual Evaluation of #LLMs for Healthcare Queries" accepted at TheWebConf 2024! 🥳 Paper: Code & Data: Blog:

2

10

65

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

Many have asked for Generative #AI and #LLM courses to learn and upskill. Here's a list of high-quality courses to binge on this pseudo-extended weekend: (1) @AndrewYNg , @awscloud , and @DeepLearningAI_ 's course on practical aspects of LLMs and deployment:

Generative AI with LLMs - DeepLearning.AI

Learn the fundamentals of how generative AI works, and how to deploy it in real-world applications. Equip yourself with the technical skills and intuition needed to succeed in the growing demand for...

www.deeplearning.ai

2

18

64

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

Work hard, but also have fun along the way 😁 End of academic year celebration with my group members. Fun night of dinner and games (pictionary!)

0

2

64

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

Please RT! Postdoc opening in my group @gtcomputing @GTCSE on exciting topics in recommendation systems #recsys and an opportunity to collaborate with our industry partners! Send in your applications by email asap! @mlatgt @IDEaSatGT @GeorgiaTech

0

27

60

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Such a pleasure to present our research on misinformation (multi-lingual, multi-modal, and multi-platform) in a plenary panel at #ICA22 ! Works funded by three @NSF grants, in collaborations w/ @dvshah @prowag @munmun10 @evilproffy & more.

3

8

58

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

Call for student scholarships to attend the The Web Conference 2021 (WWW 2021 #www2021 ). Please apply by March 5th. @TheWebConf @jure

0

29

58

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Feeling amazing after the first in-person data mining conference after 3 years! #KDD2022

1

56

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

Lynnette is a social computing Ph.D. student at CMU @CarnegieMellon and had to go through a heart transplant. Please consider donating to help with the medical expenses:

lynnette ng

@quarbby

1 year

Hello! So I disappeared because more than 40 days ago I went into end stage heart failure. I was supported by machines. By God’s Grace, I received a perfect heart and just finish a heart transplant Day 5.

15

6

153

4

17

56

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Introducing MisinfoCorrect, an open-source RL framework to generate counter-misinformation responses to misinformation! 📢New paper @TheWebConf '23 Link: This shows the promise of LLMs for societal good – here, to create a safe information ecosystem.

1

9

55

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Congratulations to my mentee Andy Chung for getting the @NSF CSGrad4US fellowship! 🎉 NSF will support 3 years of his Ph.D. Andy is a @gtcomputing alumnus and worked at Amazon for a few years before applying for Ph.D. Congrats! 👏

1

53

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

If you make hateful remarks, your followers become 5x more likely to do the same (Fig 1) #RacismIsAVirus But speaking against it discourages others from turning hateful (Fig 2). We can fight racism by simply speaking up! More on #racism during #COVID19 :

1

15

52

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

11 months

LLMs hallucinate😵‍💫more in non-English languages than in English. We create XLingEval & XLingHealth, a cross-lingual framework & a dataset to evaluate LLM performance across languages. ArXiv: Github: Blog:

1

8

51

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

6 years

News article on my research: "This AI Predicts Online Trolling Before It Happens"

This AI Predicts Online Trolling Before It Happens

Stanford's Dr. Srijan Kumar's AI is being used to weed out fake reviews, but could also help predict and curb online harassment.

www.pcmag.com

3

5

48

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

9 months

RAG reduces but does not solve hallucinations in LLMs. For a finance application use case, our team at Lighthouz AI has found that GPT-3.5 + RAG gives perfect responses only 57% times, when asked simple queries. Rest of the responses are hallucinating, incorrect, partially

8

5

52

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

7 months

Check out this new RAG eval dataset for finance applications on @huggingface :

lighthouzai/finqabench · Datasets at Hugging Face

huggingface.co

1

12

48

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

10 months

I will be at NeurIPS next week! Would love to talk about all things AI and gen AI reliability, safety, security, and trust!

1

48

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Do you use multimodal (image + text) models in your work? Have you tested the stability of your model against plausible text variations? New #EMNLP2022 work w/ @verma22gaurav shows that multimodal models can be ultra-fickle -- performance drops by 20%+

0

5

45

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

5 years

New paper @kdd_news #KDD2019 w @jure "Predicting Dynamic Embedding Trajectory in Temporal Interaction Networks". Our JODIE model generates & forecasts time-evolving embedding, scales to v large networks, & predicts temporal links & node labels. Code & data:

GitHub - claws-lab/jodie: A PyTorch implementation of ACM SIGKDD 2019 paper "Predicting Dynamic...

A PyTorch implementation of ACM SIGKDD 2019 paper "Predicting Dynamic Embedding Trajectory in Temporal Interaction Networks" - claws-lab/jodie

github.com

1

5

47

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

11 months

🧨LLMs respond better in English than in non-English languages, our latest research shows! Introducing XLingEval & XLingHealth, a cross-lingual framework & a dataset to evaluate LLM performance across languages. ArXiv: Git:

Mohit Chandra

@mohit__30

11 months

Dr. Google is a passé; enter the new “physician,” Dr. ChatGPT. But are LLMs indeed good at answering healthcare queries? How about in different languages? Where do they succeed, and where do they fail? Our new paper explores these questions 🤖🏥 1/n

3

13

55

1

8

45

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

5 years

Looking forward to welcoming everyone to the CyberSafety workshop at @TheWebConf . Keynotes from Reddit, YouTube, Boston University @gianluca_string , UIUC. Many exciting talks and papers. Come to learn how to make the web better, safe, and more inclusive!

1

5

43

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

2 days to go for the @ICWSM Data Challenge! An exciting day with 8 papers, 2 invited talks ( @informor & Kathleen Carley), and an engaging panel ( @sibel_adali @kourtellis @gvrkiran @subho_mpi @KaiShu0327 ) No registration fee to attend! More at:

1

13

44

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

Highly recommend everyone, including PhD students, researchers, and developers, to learn *streamlit*. It makes prototyping awfully easy and helps you sell your vision and idea concretely.

Andrej Karpathy

@karpathy

1 year

Carve out a few hours to learn Powerful for rapid prototyping, interactive visualization. It's a hammer and you'll start seeing a lot of nails.

111

354

4K

1

7

44

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

New #icwsm21 paper on detecting deception in group video discussions using networks: with @jure @vssubrah & Chongyang Data available at: Presentation at @icwsm next week. Tune in! @gtcomputing @GTCSE @mlatgt @IDEaSatGT

0

6

44

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

It was great to have Vibhor as a visiting PhD student at Georgia Tech! He is working on exciting research on LLMs and #AI . Stay tuned for cool publications to come from him.

Vibhor Agarwal

@vibhoragarwall

1 year

It was a memorable summer in Atlanta and a great research experience with @srijankedia , @munmun10 at @GeorgiaTech . Looking forward to continuing the collab remotely! 😊

1

0

33

2

42

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

What a cool way to conclude the week 😁 Spoke at the Stanford Trust and Safety conference about my group's research on detection models and mitigation strategies for cross-platform misinformation and platform ban evasion ✨

0

1

43

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Excited to give the inaugural talk at Google's Information Quality speaker series! 🎉 I will talk about Misinformation Detection, Fortification, Attribution, and Mitigation, covering several years of works my research group has done!

1

42

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

10 months

I’ll be at #NeurIPS2023 this week. I would love to talk about LLM reliability, RAG evals, chatbot benchmarking, GenAI agents, gen AI developer tools, open source, and more.

1

2

41

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

#NeurIPS2022 1 of 4 reviewers engaged with our rebuttal and increased score from 3 --> 6. I hope the other 3 reviewers respond too, as students have put a LOT of effort into the paper and rebuttal.

1

40

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

9 months

@karpathy It is possible/likely that the question and its solution is already in the training data. Have you tried with 'new' questions?

2

0

41

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

We all know information spreads across communities and platforms. But how? What are the typical pathways of spread? Can we predict spread? @AhrenJin answers these questions in #KDD2023 presentation! Paper: Website:

1

11

41

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

5 months

🤮 Using an LLM to evaluate another LLM sucks! LLMs hallucinate and make mistakes, so why would the judge LLM not do the same? Blindly trusting the eval LLM is worse than not doing evals at all, as it gives a false sense of trust. Even if the judge LLM is "more powerful" than

1

6

41

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

People correct those who spread online misinformation. But not everyone gets corrected. Who gets corrected (and who doesn't)? 📢 New paper at ACM @WebSciConf 2023: "Characterizing and Predicting Social Correction on Twitter" by Eric Ma, Bing He, Nathan Subrahmanian, and me

1

2

40

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

My *first* fully in-person class came to an end yesterday! The class was on online social media and networks, adversarial attacks, #misinformation , #hatespeech . Students did very creative projects on these topics too!

1

37

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

5 years

Do you work on online trust, safety, and well-being? Submit your papers to the Cybersafety workshop at #WWW2020 @TheWebConf . Details:

1

9

36

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Prof. Hima Lakkaraju @hima_lakkaraju giving an engaging talk on the interpretability and robustness of AI models at the Kavli Frontiers of Science Symposium @KavliFoundation !

1

5

37

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Great talk today by @liuhuan @SCAI_ASU at @IDEaSatGT Georgia Tech on "Social Media Mining"!

1

2

37

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

New paper "Examining the impact of sharing COVID-19 misinformation online on mental health" published at Scientific Reports @SciReports 🎉 Funded by @NSF , led by @verma22gaurav & done with @munmun10 , @TAledavood & Ankur! @gtcomputing @GTCSE @ICatGT

Gaurav Verma

@verma22gaurav

2 years

What effect does sharing misinformation online has on mental health indicators like anxiety?🤔 We find that individuals who shared COVID-19 misinformation experienced exacerbated anxiety when compared to similar individuals who didn't! @SciReports 🧵👇1/

1

4

27

1

35

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

I am delighted to present my group's research work next week at IIM Ahmedabad, India's topmost Business School!

IIM Ahmedabad

@IIMAhmedabad

3 years

The @IIMA_RP is hosting a webinar in collaboration with IIMA Brij Disa Centre for Data Science and Artificial Intelligence on Advances in AI for Social Cyber-Safety by Prof. Srijan Kumar from Georgia Institute of Technology. For registration, email us at respub @iima .ac.in

0

4

15

1

34

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

10 months

😍Eventful day discussing #AI safety, trust, reliability, and AI’s impact on the society! The role of AI in our personal and professional lives is rapidly increasing. AI’s impact on the society is huge already. As AI is being used to drive important decisions for us, the

2

1

34

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

5 years

Slides for my #KDD2019 presentation "Predicting Dynamic Embedding Trajectory in Temporal Interaction Networks" with @jure

0

7

32

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

Non-expert users are responsible for 96% (!!) of all counter misinformation tweets. More in our paper "The Role of the Crowd in Countering Misinformation: A Case Study of the COVID-19 Infodemic" presented this week at IEEE BigData 2020 @ieeebigdata Link:

0

6

33

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

Congrats to the students Bing He, Caleb Ziems, Sandeep Soni and collaborators @Diyi_Yang and @profnaren for acceptance of the paper on #HateSpeech and counter-speech during the COVID-19 pandemic at the #ASONAM21 conference 🎉 @ASONAM_News

0

2

32

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

🎉Our paper selected for spotlight talk at the #NeurIPS2022 Temporal Graph Learning workshop! Congratulations to @ksartik @RohitSridhar8 @rstriv 👏 Paper: Imperceptible Adversarial Attacks on Discrete-Time Dynamic Graph Models PDF coming soon

0

3

32

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

Deep learning models are used to classify entities (detect malicious or at-risk users, etc). But are the models vulnerable to adversaries? Yes, very! More in our #KDD2021 paper, where we create a personalized text generation attack PETGEN. Code & data:

1

6

33

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

My Ph.D. student @ksartik presenting his paper at @NeurIPSConf Temporal Graph Learning workshop. The slides and paper link are below 👇

1

0

33

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

🚨New paper alert🚨 "The Role of the Crowd in Countering Misinformation: A Case Study of the COVID-19 Infodemic" with my students to be presented next week at IEEE Big Data 2020! Link: We investigate how non-expert people fight misinfo on Twitter.

1

2

33

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

6 years

How do communities fight? Find out at paper presention "Community Interaction and Conflict on the Web" happening at noon in Bellacour 2. Web and Society Track. #TheWebConf #WWW2018 w/ @jure @williamleif @jurafsky

0

7

32

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

I am giving a talk at the IEEE Computer Society San Diego chapter on detection and mitigation of online harms using NLP, graph, and adversarial ML methods! Topic: Advances in AI for Web Integrity, Equity, and Well-Being Time: 5:30 pm PT Registration is free. Link below.

2

32

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

As an assistant professor, your time is usually split as... 50% supervising students 50% writing papers 50% writing grants 50% teaching 50% PR 50% recruiting 50% giving talks 50% HR 50% service 50% everything else

Andrew Gazdecki

@agazdecki

2 years

As a startup founder your time is usually split up as... 50% sales 50% marketing 50% product 50% operations 50% support 50% strategy 50% recruiting 50% everything else

121

289

2K

2

3

32

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

11 months

📣New paper: "Representation Learning in Continuous-Time Dynamic Signed Networks" at @cikm2023 ! Code: @ksartik and Mohit presented the paper this week! @GTCSE

GitHub - claws-lab/semba: Representation Learning in Continuous-Time Dynamic Signed Networks, CIKM...

Representation Learning in Continuous-Time Dynamic Signed Networks, CIKM 23 - claws-lab/semba

github.com

1

4

31

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

I had a great time chatting with Dr. Chirigati from Nature Computational Science about myself and my group's work on Anti-Asian hate, misinformation, and web safety and integrity! Read the Q&A piece below!

Nature Computational Science

@NatComputSci

2 years

To celebrate Asian American and Pacific Islander heritage month, we talked to Dr. Srijan Kumar ( @srijankedia ) about his work on combating online malicious activities and actors, including misinformation and anti-Asian hate speech. #AAPIHM

0

7

29

0

1

32

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

5 years

"Higher-Order Label Consistency and Spreading in Graphs", with Dhivya Eswaran and Christos Faloutsos @bear5claw , is accepted to #www2020 @TheWebConf ! New method for belief propagation with higher-order structures. Applications in web, social, finance, & cybersecurity domains.

1

31

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Ban evasion poses the biggest threats to online safety/integrity. It is trivial for bad actors to create new account after being banned! We did the first work on ban evasion detection, prediction, and attribution. Models work well: Talk tmrw @TheWebConf

2

5

31

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

3 years

Congrats to Sejoon on his first PhD paper accepted at @CIKM2021 🎉🎉 "Influence-guided Data Augmentation for Neural Tensor Completion" Link: Sejoon has created a new data augmentation technique for recommender systems Code + data:

3

1

31

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

1 year

Excited to present *two* #keynotes at ML for Finance and Robust NLP for Finance workshops at #KDD2023 on "Robustness of LLMs and Multimodal Models"! Slides here: I will highlight four recent research papers on:

Login to your Canva account

Login to access and create your next design.

www.canva.com

1

2

30

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

10 months

🚀 Many have asked for Generative AI courses and hands-on practice material. Here are 12 learning resources: 👉1. Beginner’s Guide to GenAI: LLMs, RAG, and implementation practice: 👉2. How to Build LLM Apps that can See Hear Speak using OpenAI Whisper

2

5

28

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

2 years

Graph Neural Networks are revolutionizing many domains, including Web Trust, Safety, and Integrity! Excited to present my group's work at Stanford's Graph Learning Workshop 🤩Great to be back at Stanford!

0

2

29

Srijan Kumar - Lighthouz AI (YC S24) 🤖

@srijankedia

4 years

Congratulations to my PhD student, Sejoon Oh, for being selected as a 2021 Twitch Research Fellowship Finalist! @GTCSE @mlatgt @gtcomputing @GeorgiaTech @chaoukita

GaTech CSE

@GTCSE

4 years

🎉Congratulations to CSE Ph.D. student Sejoon Oh for being selected as a finalist for the 2021 @Twitch Research Fellowship! Sejoon is advised by @srijankedia and his research focuses on user modeling and recommender systems. Read more:

0

1

13

0

29