Aadirupa Saha @AadirupaSaha Twitter profile

Pinned Tweet

Aadirupa Saha

1 month

PhD Research Intern @ #Apple ! Focus: Personalization, preference-based learning, RLHF, LLM etc. Prefer candidates with strong ML/RL/OL theory backgrounds w/ coding skills Apply - Pls fwd in your circles! #MachineLearning #LLM #RL #neurips #ICLR #ICML

This link will take you to a page that’s not on LinkedIn

lnkd.in

5

69

402

Last Seen Profiles

@iusedtohavecats

@norsedog

@NitroElFamous

@IanFich

@GuiltyKilt_

@BinorRaja

@cukienaknikmati

@Goddessmia190

@annabunney

@inicessaaa

@Da33672Darcy

@nudefreebi

@marin_blqtokyo

@HauseRose63839

@yuuyuuandyuu

@chapati_lions

@4ever_mor

@SRFlubber

@RUFieldHockey

@daddybears1349

@navyily

@bokeplokalmalam

@MTauseef707

@NightOwlNSFW

@Jiz

@miro7280

@richsamurai_flr

@France3CA

@EstreaWilliams

@sumire06_sumire

@marineinstitute

@vixendeeper

@Mikey_Sul

@mint_fps

@amyfclapham

Aadirupa Saha

@AadirupaSaha

2 years

I’m happy to share that I’m starting a new position as Research And Development Scientist at Apple!

10

1

205

Aadirupa Saha

@AadirupaSaha

9 months

Glad we got two papers at #ICLR2024 1. Variance adaptive Bandits () w/ ever-amazing and fun-to-work Brano! 2. Bandits Meet Mechanism Design () w/ Thomas, @haifengxu0 and Christos. Thanks to Thomas we actually landed a spotlight!

Bandits Meet Mechanism Design to Combat Clickbait in Online Recommendation

We study a strategic variant of the multi-armed bandit problem, which we coin the strategic click-bandit. This model is motivated by applications in online recommendation where the choice of...

arxiv.org

1

6

43

Aadirupa Saha

@AadirupaSaha

6 months

If you are at #AISTATS24 , please stop by our papers: 1. Faster Convergence with Multiway Preferences: on Optimization w/ relative feedback 2. Think Before You Duel on Understanding Complexities of Preference Learning under Budgets

Think Before You Duel: Understanding Complexities of Preference...

We consider the problem of reward maximization in the dueling bandit setup along with constraints on resource consumption. As in the classic dueling bandits, at each round the learner has to...

arxiv.org

0

3

32

Aadirupa Saha

@AadirupaSaha

11 months

I'm so caught in a whirlwind of excitement and anxiety. Omg we are less than 2 weeks away and it's my first tutorial at NeurIPS! If you find yourself in New Orleans on Dec 11, swing by Hall C2 at 1:45pm , bet you won't regret! #NeurIPS2023 #EmotionOverload

Aditya Gopalan

@today_itself

11 months

Join @AadirupaSaha and me at #NeurIPS2023 for our tutorial "Do You Prefer Learning with Preferences?" on 11 Dec! Our program is packed with insights about how to make sense of preference data in ML, build prediction models, draw inferences, and useful applications. (1/2)

3

8

51

0

2

22

Aadirupa Saha

@AadirupaSaha

1 year

Save the dates & pls fwd! The workshop we have been waiting for is here: NEW FRONTIERS IN FEDERATED LEARNING. Sept 8,9 @ TTIC. Excellent line of speakers, invited talks, research breakout sessions, open problems! Anything else? Feel free to suggest!

Home

Federated Learning has garnered much attention recently with its promises to extract insights from sensitive client data without compromising privacy, thus allowing for privacy-preserving collabora...

sites.google.com

2

8

22

Aadirupa Saha

@AadirupaSaha

5 months

To all my Bandit folks attending #ICLR2024 this time, let us know your thoughts on our work: Variance Adaptive Thompson Sampling (w/ Branislav)! #iclr24 #ICLR2024 -- Why regret if you are already confident. Only Pay for What Is Uncertain! Details 👇

Only Pay for What Is Uncertain: Variance-Adaptive Thompson Sampling

Most bandit algorithms assume that the reward variances or their upper bounds are known, and that they are the same for all arms. This naturally leads to suboptimal performance and higher regret...

arxiv.org

1

3

15

Aadirupa Saha

@AadirupaSaha

3 months

Gentle reminder for #ICML24 attendees! Don't miss our workshop: Models of Human Feedback for AI Alignment on Fri July 26 We've got an exciting lineup of talks + received overwhelming #submissions . Pls stop by to chat with the authors at the poster! #AI #ML #Alignment #RLHF #NLP

Aadirupa Saha

@AadirupaSaha

5 months

Attention Alignment enthusiasts! Mark the dates & pls share. The wrkshp we've been waiting is finally here Focus: Role of appropriate modeling of human feedback before training alignment algos. Surprisingly(!) the topic is mostly neglected so far, but no more #ICML #AI #LLM #GPT

0

5

0

15

Aadirupa Saha

@AadirupaSaha

1 month

Also send your resume, website, and 1 or 2 representative papers to mlr-ppbl-internship @group .apple.com and CC me! Apply soon!

1

0

13

Aadirupa Saha

@AadirupaSaha

2 years

1. Why regret if you are confident? Only Pay for What Is Uncertain! Check out our new paper on variance-adaptive Bayesian MAB for unknown heterogeneous reward variances.

Only Pay for What Is Uncertain: Variance-Adaptive Thompson Sampling

Most bandit algorithms assume that the reward variances or their upper bounds are known, and that they are the same for all arms. This naturally leads to suboptimal performance and higher regret...

arxiv.org

1

13

Aadirupa Saha

@AadirupaSaha

1 year

Fun starts tomorrow @TTIC_Connect Join us for a two-day workshop on Federated Learning. Invited talks, breakout sessions, open problems, panel discussion and ... good times! See you there!

Home

Federated Learning has garnered much attention recently with its promises to extract insights from sensitive client data without compromising privacy, thus allowing for privacy-preserving collabora...

sites.google.com

1

2

11

Aadirupa Saha

@AadirupaSaha

2 years

I am so excited to organize a workshop on "The Many Facets of Preference-based Learning" at the International Conference of Machine Learning (ICML), 2023, Hawaii, with my amazing co-organizers: - Viktor Bengs (LMU, Germany) - Rober…

Aadirupa Saha on LinkedIn: MFPL-ICML-2023

I am so excited to organize a workshop on "The Many Facets of Preference-based Learning" at the International Conference of Machine Learning (ICML), 2023…

www.linkedin.com

0

1

10

Aadirupa Saha

@AadirupaSaha

2 years

Excited to announce our ICML workshop on Complex Feedback in Online Learning ICML'22 Workshop July 23, 2022, Baltimore USA Inviting your contributions! Submission details are on the website. Deadline May 27. Submit your best work! See you at ICML 🙂

0

3

9

Aadirupa Saha

@AadirupaSaha

1 year

Anyone attending AIStats this year, please stop by our posters : ) We have three papers accepted - 1. ANACONDA: Improved Dynamic Regret Algorithm for Adaptive Non-Stationary Dueling Bandits (with Thomas Kleine Buening): 2. One Ar…

Aadirupa Saha on LinkedIn: Anyone attending AIStats this year, please stop by our posters : ) We...

Anyone attending AIStats this year, please stop by our posters : ) We have three papers accepted - 1. ANACONDA: Improved Dynamic Regret Algorithm for Adaptive…

www.linkedin.com

0

8

Aadirupa Saha

@AadirupaSaha

7 months

Agreed! I suffer this often including #icml2024 Is there no way we can hv a stable review process w/ proper accountability? We're subjected to 2 unreasonable+reluctant reviewers-Me spending 3+days rebutting, sending notes to AC/SACs,no resolution yet! Any other option we've here?

Gergely Neu

@neu_rips

7 months

would anyone be interested in a follow-up rant from the reviewers' side? i have built up plenty of material for that during the last 2.5 days that i've spent entirely on responding to rebuttals...

6

0

36

0

7

Aadirupa Saha

@AadirupaSaha

2 years

CALL FOR PAPERS!!! Excited to organize our new workshop on "The Many Facets of Preference-based Learning" at this year's ICML, Hawaii, with my amazing co-organizers @BengsViktor , Robert Busa-Fekete, Mohammad Ghavamzadeh, and Branislav Kveton. Website:

MFPL-ICML-2023

WORKSHOP Overview

sites.google.com

1

3

7

Aadirupa Saha

@AadirupaSaha

5 months

Attention Alignment enthusiasts! Mark the dates & pls share. The wrkshp we've been waiting is finally here Focus: Role of appropriate modeling of human feedback before training alignment algos. Surprisingly(!) the topic is mostly neglected so far, but no more #ICML #AI #LLM #GPT

MHFAIA@ICML2024

@mhf_icml2024

5 months

Announcing the Workshop on Models of Human Feedback for AI Alignment at ICML 2024 ( @icmlconf )! Check out the website for the call for papers and more details. Website: Submission Deadline: May 31, 2024

1

5

9

0

5

Aadirupa Saha

@AadirupaSaha

2 years

Excited about our workshop at TTIC this summer - July 13-15th on "Online decision-making and real-world applications". We invite graduate students to present their work in the poster session (Deadline: June 30th). We also invite researchers to present interesting open problems!

Microsoft Research

@MSFTResearch

2 years

Organized by researchers from Microsoft Research's NYC lab, TTIC Chicago Summer Workshop will look at the contemporary decision-making models that are being actively researched, aiming to bridge theory and practice. Register by the first week of July:

1

4

11

1

3

7

Aadirupa Saha

@AadirupaSaha

5 months

Tired of overpriced deceptive merchandise and misleading posts on online marketplaces? Check our #ICLR2024 paper for a remedy! with @thomasklbg , @haifengxu0 and @olethros . More details in the thread below:

Bandits Meet Mechanism Design to Combat Clickbait in Online Recommendation

We study a strategic variant of the multi-armed bandit problem, which we coin the strategic click-bandit. This model is motivated by applications in online recommendation where the choice of...

arxiv.org

Aadirupa Saha

@AadirupaSaha

9 months

Glad we got two papers at #ICLR2024 1. Variance adaptive Bandits () w/ ever-amazing and fun-to-work Brano! 2. Bandits Meet Mechanism Design () w/ Thomas, @haifengxu0 and Christos. Thanks to Thomas we actually landed a spotlight!

1

6

43

2

0

6

Aadirupa Saha

@AadirupaSaha

5 months

Gentle reminder: CFP is open till May 31 Don't miss out! Topics: IRL, Imitation Learning, RLHF, LLM, AI Alignment, AI Safety, Robotics, Pref Learning, Social Choice, OR, Cognitive Sc, Behavioral Econ & more. #Research #AI #ICML2024 #LLMs #chatgpt4 #ML

Models of Human Feedback for AI Alignment ICML 2024 - Call for Papers

TLDR: Length: 4-8 pages (not including references and appendix) Formatting: ICML latex style. Please change the footnote to reflect the name of the workshop. Dual Submission Policy: We accept...

sites.google.com

MHFAIA@ICML2024

@mhf_icml2024

5 months

Announcing the Workshop on Models of Human Feedback for AI Alignment at ICML 2024 ( @icmlconf )! Check out the website for the call for papers and more details. Website: Submission Deadline: May 31, 2024

1

5

9

0

1

6

Aadirupa Saha

@AadirupaSaha

1 year

The deadline is less than a month away! Don't forget to submit your work : ) Questions? -- Just shoot an email @learningpreference @gmail .com Details here -- #ICML2023

Aadirupa Saha

@AadirupaSaha

2 years

CALL FOR PAPERS!!! Excited to organize our new workshop on "The Many Facets of Preference-based Learning" at this year's ICML, Hawaii, with my amazing co-organizers @BengsViktor , Robert Busa-Fekete, Mohammad Ghavamzadeh, and Branislav Kveton. Website:

1

3

7

1

6

Aadirupa Saha

@AadirupaSaha

8 months

Hope it'll be useful for the community, that’s all we need to keep going! If u can spare a few mins & leave any feedback/correction/content suggestions, we owe u🙂 & a blog acknowledgment! So pls be critical! + Do check out @BengsViktor 's amazing survey

Preference-based Online Learning with Dueling Bandits: A Survey

In machine learning, the notion of multi-armed bandits refers to a class of online learning problems, in which an agent is supposed to simultaneously explore and exploit a given set of choice...

arxiv.org

Aadirupa Saha

@AadirupaSaha

8 months

@BengsViktor is still mad as we’ve been dragging it forever-all thanks to my procrastination skills! But we’re finally ready with our 1st blog on “Preference Learning”-We both have been working in the area for a while + Glad to have Eyke on board with us!

0

2

3

1

5

Aadirupa Saha

@AadirupaSaha

1 year

Last call for papers!! Deadline extended to June 16 AOE! Notification: June 26 Workshop: July 28 @ #ICML23 Accepted papers will be presented as posters & some as contributed talks Email: learningpreference @gmail .com Please consider submitting!

MFPL-ICML-2023

WORKSHOP Overview

sites.google.com

Aadirupa Saha

@AadirupaSaha

2 years

CALL FOR PAPERS!!! Excited to organize our new workshop on "The Many Facets of Preference-based Learning" at this year's ICML, Hawaii, with my amazing co-organizers @BengsViktor , Robert Busa-Fekete, Mohammad Ghavamzadeh, and Branislav Kveton. Website:

1

3

7

0

1

5

Aadirupa Saha

@AadirupaSaha

1 year

Please consider registering if you are planning to be around TTI Chicago during Sept 8-9th.

TTIC

@TTIC_Connect

1 year

@McAllesterDavid @YishayMansour New Frontiers in Federated Learning aims to underline the limitations of optimization and privacy aspects in #federatedlearning . Learn more and register for this workshop:

1

0

2

0

5

Aadirupa Saha

@AadirupaSaha

14 days

I know people are busy with upcoming deadlines but would you mind leaving a quick opinion about #NeurIPS2024 reviewing process from your recent experience? I am really curious! #NeurIPS #ICLR2025 #ML #LLMs #AI #MachineLearning #deeplearning #Research #PeerReview

Couldn't be happier!

11

Satisfied

37

Disappointed

48

Opaque and unfair!

27

0

1

5

Aadirupa Saha

@AadirupaSaha

8 months

Also forgot to mention but BIG thanks to @bremen79 and @LattimoreTor for guiding us through the initial logistical hurdles and all the suggestions to kickstart. Goes without saying, we’re much inspired by & Thanks! : )

Bandit Algorithms

banditalgs.com

Aadirupa Saha

@AadirupaSaha

8 months

@BengsViktor is still mad as we’ve been dragging it forever-all thanks to my procrastination skills! But we’re finally ready with our 1st blog on “Preference Learning”-We both have been working in the area for a while + Glad to have Eyke on board with us!

0

2

0

4

Aadirupa Saha

@AadirupaSaha

7 months

@neu_rips @neu_rips Why can't authors and reviewers give scores to each other? We can aggregate the rating scores of an individual across all their conf. profiles & those ratings should matter for u to be a qualified reviewer? This should hold us accountable to review properly? #icml2024

1

0

4

Aadirupa Saha

@AadirupaSaha

8 months

(1/2) If you've ever encountered MNL Assortment-Opt problem, the go-to approach is to offer the same set of products repeatedly until your customer is really annoyed and selects no item! In fact, it requires a "belief" that no selection is their most preferred choice :-( Oh no!

1

0

3

Aadirupa Saha

@AadirupaSaha

2 years

2. Our variance-sensitive regret analysis shows improved instance-dependent learning rates in the low-variance regimes, unlike standard MAB algorithms that pay for worst-case variance, and of course end up regretting a lot more!

0

3

Aadirupa Saha

@AadirupaSaha

2 years

Please go over the CFP and submit your best work: DATES: Submission Deadline: June 5, 2023 Notification of Acceptance: June 26, 2023 Workshop: July 28 (Friday) or 29 (Saturday), 2023 Questions? Feel free to reach us @learningpreference @gmail .com

0

1

3

Aadirupa Saha

@AadirupaSaha

1 year

@thegautamkamath @JubaZiani I learned so much from that course too, thanks a lot for putting it up.

0

3

Aadirupa Saha

@AadirupaSaha

2 years

There is still time to submit your posters and open problems to the workshop. The deadline (June 30th) is approaching fast! Also do not forget to register here if you are planning to attend: Hope to see you in July! Looking forward to it.

Registration for 2022 TTIC Summer Workshop: New Models in Online Decision Making for Real-World...

for

docs.google.com

BeyondRL TTIC

@BeyondrlT

2 years

🤖🤖 Announcement: Open problem session about interactive decision-making 🤖🤖 We invite researchers👩‍🎓🧑‍🎓 to present interesting open problems in a designated session at the workshop Lets stir the discussion!

2

1

2

0

1

3

Aadirupa Saha

@AadirupaSaha

2 years

We are organizing a workshop at TTIC, Chicago, this summer! July 13-15th 2022. Join us in person or virtually to discuss more.

Home

WORKSHOP Overview

sites.google.com

0

2

Aadirupa Saha

@AadirupaSaha

7 months

@neu_rips #icml2024 There shld be some accountability from the #Reviewers or they hv all the power to do anything while the onus is on the authors to bow down no matter how unreasonable & reluctant the other end is? I wrote to PCs & SAC again for more revs but nothing yet, what else to do?

0

2

Aadirupa Saha

@AadirupaSaha

1 year

Pretty excited about this and definitely planning to attend in person. If you are around UChicago and want to discuss any problem/ just chat, please ping me :)

TTIC

@TTIC_Connect

1 year

DavidFest and YishayFest will be held on August 17-18. These celebratory/milestone birthday workshops are in honor of TTIC’s @McAllesterDavid , and friend-of-TTIC @YishayMansour . Learn more and register:

1

0

1

0

2

Aadirupa Saha

@AadirupaSaha

1 year

Register here: Can't wait to see you all!

Registration for TTIC Summer Workshop on New Frontiers in Federated Learning (Sept 8-9, 2023)

Form to register for attending the TTIC Summer Workshop on New Frontiers in Federated Learning, September 8-9, 2023

docs.google.com

0

2

Aadirupa Saha

@AadirupaSaha

1 year

@neu_rips @NeurIPSConf @icmlconf @iclr_conf I totally agree! Drained me both as an AC and author. A decision of an 8pg paper should not require another discussion of 8+ pages. This encourages authors to submit half-polished papers and later fix things on demand. Also unfair to those where reviewers are not engaging at all.

0

2

Aadirupa Saha

@AadirupaSaha

1 year

It was so much fun!

TTIC

@TTIC_Connect

1 year

Today is the start of YishayFest and DavidFest, as well as our 2023 Summer Workshop Program! Learn more about our upcoming workshops:

0

5

0

2

Aadirupa Saha

@AadirupaSaha

1 year

#AISTATS2023 Our poster is today at 4:30 CET ( #78 ): ANACONDA: An Improved Dynamic Regret Algorithm for Adaptive Non-Stationary Dueling Bandits Thomas (my co-author) will be at the poster. Please stop by and ask as many questions as you have - Thomas is always up for questions!

4

0

2

Aadirupa Saha

@AadirupaSaha

8 months

@BengsViktor is still mad as we’ve been dragging it forever-all thanks to my procrastination skills! But we’re finally ready with our 1st blog on “Preference Learning”-We both have been working in the area for a while + Glad to have Eyke on board with us!

0

2

Aadirupa Saha

@AadirupaSaha

1 year

For students: We still have some funding left. (Our budget is limited, first come first serve and the reimbursement amount would vary depending on the total participation). Contact us @ flworkshop23 @ttic .edu

0

Aadirupa Saha

@AadirupaSaha

1 year

Fair learning could be vulnerable to a small amount of malicious noise, esp. in imbalanced group sizes. But we show how randomization can come to rescue for several notions of fairness: Parity, equal opportunity, Equalized Odds, Calibration. Learn more @

0

2

Aadirupa Saha

@AadirupaSaha

3 months

Exciting symposium in the Bay Area! This is a fantastic opportunity to engage with leading industry experts and explore cutting-edge research. Please consider submitting an abstract if you are around and planning to join in person!

Jerremy Holland

@jerremy

3 months

The BayLearn 2024 abstract submission deadline has been extended to Aug 5, 2024. If you are working on cutting-edge ML research, please consider submitting an abstract. BayLearn 2024 will be an in-person event, held on Oct 10, 2024:

0

5

7

0

2

Aadirupa Saha

@AadirupaSaha

1 month

@tinner_he Can you please email me your resume/interests along with any special requirements you might have? We can take it from there.

0

1

Aadirupa Saha

@AadirupaSaha

1 year

@mlittmancs Happy birthday!

0

1

Aadirupa Saha

@AadirupaSaha

1 year

Can't agree more!

Anand Bhattad

@anand_bhattad

1 year

Reviewing for #NeurIPS2023 has become a mental marathon. The constant AC nudges & endless engagement is overwhelming. Every response leads to more emails. It's not just me; many reviewers share the sentiment. I'd prefer a longer rebuttal rather than this non-stop back-and-forth😓

5

0

19

0

1

Aadirupa Saha

@AadirupaSaha

5 months

@mhf_icml2024 @icmlconf Submission deadline: May 31st AOE Acceptance notification: June 17th Workshop: July 26th Related Topics: Submission Portal: Workshop includes invited talks (great line of speakers!), panel discussions, open problems, and more!

ICML 2024 Workshop MFHAIA

Welcome to the OpenReview homepage for ICML 2024 Workshop MFHAIA

openreview.net

0

Aadirupa Saha

@AadirupaSaha

5 months

We handle Bayesian Multiarmed bandits with known and unknown heterogeneous reward variances with Thompson Sampling. Our analysis captures a "cute tradeoff" between regret & reward variances --- showing faster learning in low-variance regimes. + Range of expts got us covered!

0

1

Aadirupa Saha

@AadirupaSaha

7 months

@neu_rips Already flagged them to AC and SAC but no response yet. Meanwhile, another R2 (out of remaining 3) simply dint care about the rebuttal & says "Sorry for keeping my scores" based on some completely misleading conclusion out of 1.5pg rebuttal we wrote for R2. How is this justified?

0

1

Aadirupa Saha

@AadirupaSaha

1 year

@SOURADIPCHAKR18 Thanks!

0

1

Aadirupa Saha

@AadirupaSaha

5 months

We proposed a strategic click-bandit model for online RecSys, where arms (merchandise) can be strategically manipulated to maximize their CTR. But we give an incentive-aware algorithm that ensures truthful behavior and minimizes regret across all approximate Nash equilibria!

0

Aadirupa Saha

@AadirupaSaha

1 year

@chijinML Congrats!!

0

Aadirupa Saha

@AadirupaSaha

6 months

Nice work!

Bogdan Mazoure

@bogdan_mazoure

6 months

Our team @Apple will be in Vienna next week for #ICLR2024 , where we will be presenting our work on using large language models as RL policies. Come drop by our poster! Website: Joint work with @max_a_schwarzer , @harsh_092 , @alexttoshev and others

1

15

71

0

1

Aadirupa Saha

@AadirupaSaha

2 years

Thanks everyone.

0

1