NYU Center for Data Science
@nyudatascience.bsky.social
Official account of the NYU Center for Data Science, the home of the Undergraduate, Master’s, and Ph.D. programs in data science. cds.nyu.edu
CDS Silver Professor Julia Kempe (@kempelab.bsky.social) is co-organizing this year's ICLR 2026 Workshop on New Frontiers in Associative Memory.
The workshop is accepting submissions related to associative memory until Feb 14.
nfam2026.amemory.net
The workshop is accepting submissions related to associative memory until Feb 14.
nfam2026.amemory.net
Workshop on Associative Memory @ICLR 2025
New Frontiers in Associative Memories Workshop @ICLR 2025. Discuss the latest multidisciplinary developments in Associative Memory. Explore synergies between Associative Memories and LLMs.
nfam2026.amemory.net
January 30, 2026 at 7:50 PM
CDS Silver Professor Julia Kempe (@kempelab.bsky.social) is co-organizing this year's ICLR 2026 Workshop on New Frontiers in Associative Memory.
The workshop is accepting submissions related to associative memory until Feb 14.
nfam2026.amemory.net
The workshop is accepting submissions related to associative memory until Feb 14.
nfam2026.amemory.net
Reposted by NYU Center for Data Science
Verifiers are increasingly being used today in RL to provide rewards. We did a systematic study on when it is the best to use LLMs to verify solutions. Check out the blog post below to learn more.
Do stronger LLMs make better verifiers? Not necessarily when grading themselves.
New work led by Courant PhD student @jacklu-me.bsky.social and CDS Asst Prof @mengyer.bsky.social shows that cross-family verification outperforms self-verification.
nyudatascience.medium.com/study-reveal...
New work led by Courant PhD student @jacklu-me.bsky.social and CDS Asst Prof @mengyer.bsky.social shows that cross-family verification outperforms self-verification.
nyudatascience.medium.com/study-reveal...
Study Reveals AI Models Are Biased Toward Solutions That Resemble Their Own Reasoning
Having AI models check the work of models from different “families” is far more effective than having them verify their own outputs.
nyudatascience.medium.com
January 30, 2026 at 6:40 PM
Verifiers are increasingly being used today in RL to provide rewards. We did a systematic study on when it is the best to use LLMs to verify solutions. Check out the blog post below to learn more.
Do stronger LLMs make better verifiers? Not necessarily when grading themselves.
New work led by Courant PhD student @jacklu-me.bsky.social and CDS Asst Prof @mengyer.bsky.social shows that cross-family verification outperforms self-verification.
nyudatascience.medium.com/study-reveal...
New work led by Courant PhD student @jacklu-me.bsky.social and CDS Asst Prof @mengyer.bsky.social shows that cross-family verification outperforms self-verification.
nyudatascience.medium.com/study-reveal...
Study Reveals AI Models Are Biased Toward Solutions That Resemble Their Own Reasoning
Having AI models check the work of models from different “families” is far more effective than having them verify their own outputs.
nyudatascience.medium.com
January 30, 2026 at 4:41 PM
Do stronger LLMs make better verifiers? Not necessarily when grading themselves.
New work led by Courant PhD student @jacklu-me.bsky.social and CDS Asst Prof @mengyer.bsky.social shows that cross-family verification outperforms self-verification.
nyudatascience.medium.com/study-reveal...
New work led by Courant PhD student @jacklu-me.bsky.social and CDS Asst Prof @mengyer.bsky.social shows that cross-family verification outperforms self-verification.
nyudatascience.medium.com/study-reveal...
CDS seeks a new Chair to lead the Center’s expansion within the Courant Institute School.
The incoming Chair will have a mandate to drive growth and oversee faculty recruitment.
The deadline for full consideration is Feb 1.
apply.interfolio.com/179190
The incoming Chair will have a mandate to drive growth and oversee faculty recruitment.
The deadline for full consideration is Feb 1.
apply.interfolio.com/179190
January 29, 2026 at 7:53 PM
CDS seeks a new Chair to lead the Center’s expansion within the Courant Institute School.
The incoming Chair will have a mandate to drive growth and oversee faculty recruitment.
The deadline for full consideration is Feb 1.
apply.interfolio.com/179190
The incoming Chair will have a mandate to drive growth and oversee faculty recruitment.
The deadline for full consideration is Feb 1.
apply.interfolio.com/179190
CDS' Pascal Wallisch shares words of wisdom for data science students.
Advice:
✅ Build a community
✅ Take advantage of the open environment
✅ Maintain a broad mindset with epistemic humility
Applications to our MS in data science are open. Apply by Feb 14:
cds.nyu.edu/admissions/m...
Advice:
✅ Build a community
✅ Take advantage of the open environment
✅ Maintain a broad mindset with epistemic humility
Applications to our MS in data science are open. Apply by Feb 14:
cds.nyu.edu/admissions/m...
January 29, 2026 at 4:09 PM
CDS' Pascal Wallisch shares words of wisdom for data science students.
Advice:
✅ Build a community
✅ Take advantage of the open environment
✅ Maintain a broad mindset with epistemic humility
Applications to our MS in data science are open. Apply by Feb 14:
cds.nyu.edu/admissions/m...
Advice:
✅ Build a community
✅ Take advantage of the open environment
✅ Maintain a broad mindset with epistemic humility
Applications to our MS in data science are open. Apply by Feb 14:
cds.nyu.edu/admissions/m...
The U.S. once produced 88% of all remote sensing research. Today, it produces less than 9%, while China now produces nearly half of all global research in the field.
nyudatascience.medium.com/china-now-do...
nyudatascience.medium.com/china-now-do...
China Now Dominates Remote Sensing Research as U.S. Funding Wanes
The U.S. has lost its lead in remote sensing technology to China, dropping from 88% of global research output to less than 9%.
nyudatascience.medium.com
January 28, 2026 at 5:19 PM
The U.S. once produced 88% of all remote sensing research. Today, it produces less than 9%, while China now produces nearly half of all global research in the field.
nyudatascience.medium.com/china-now-do...
nyudatascience.medium.com/china-now-do...
Reposted by NYU Center for Data Science
So excited that our paper got accepted to ICLR!! Check it out 👇
Can LLMs evolve human-like semantic categories?
CDS-affiliated @nogazs.bsky.social and PhD student Nathaniel Imel show that, via simulated cultural transmission, LLMs reorganize color categories toward efficient compression.
🔗 arxiv.org/abs/2509.08093
CDS-affiliated @nogazs.bsky.social and PhD student Nathaniel Imel show that, via simulated cultural transmission, LLMs reorganize color categories toward efficient compression.
🔗 arxiv.org/abs/2509.08093
January 27, 2026 at 6:33 PM
So excited that our paper got accepted to ICLR!! Check it out 👇
Can LLMs evolve human-like semantic categories?
CDS-affiliated @nogazs.bsky.social and PhD student Nathaniel Imel show that, via simulated cultural transmission, LLMs reorganize color categories toward efficient compression.
🔗 arxiv.org/abs/2509.08093
CDS-affiliated @nogazs.bsky.social and PhD student Nathaniel Imel show that, via simulated cultural transmission, LLMs reorganize color categories toward efficient compression.
🔗 arxiv.org/abs/2509.08093
January 27, 2026 at 4:57 PM
Can LLMs evolve human-like semantic categories?
CDS-affiliated @nogazs.bsky.social and PhD student Nathaniel Imel show that, via simulated cultural transmission, LLMs reorganize color categories toward efficient compression.
🔗 arxiv.org/abs/2509.08093
CDS-affiliated @nogazs.bsky.social and PhD student Nathaniel Imel show that, via simulated cultural transmission, LLMs reorganize color categories toward efficient compression.
🔗 arxiv.org/abs/2509.08093
Tuning AI models no longer needs to rely on expensive guesswork.
Courant PhD student Nick Lourie, CDS Professor @kyunghyuncho.bsky.social, and CDS Assoc. Prof. He He reveal an important new statistical tool to estimate a model's best possible performance.
nyudatascience.medium.com/taking-the-g...
Courant PhD student Nick Lourie, CDS Professor @kyunghyuncho.bsky.social, and CDS Assoc. Prof. He He reveal an important new statistical tool to estimate a model's best possible performance.
nyudatascience.medium.com/taking-the-g...
Taking the Guesswork Out of AI Training: Hyperparameter Landscapes Are Simpler Than We Thought
Courant PhD student Nick Lourie shows a method that makes choosing hyperparameters more predictable and less reliant on trial and error.
nyudatascience.medium.com
January 23, 2026 at 3:41 PM
Tuning AI models no longer needs to rely on expensive guesswork.
Courant PhD student Nick Lourie, CDS Professor @kyunghyuncho.bsky.social, and CDS Assoc. Prof. He He reveal an important new statistical tool to estimate a model's best possible performance.
nyudatascience.medium.com/taking-the-g...
Courant PhD student Nick Lourie, CDS Professor @kyunghyuncho.bsky.social, and CDS Assoc. Prof. He He reveal an important new statistical tool to estimate a model's best possible performance.
nyudatascience.medium.com/taking-the-g...
The CDS Capstone Project course held its Final Poster Presentations at the Rosenthal Pavilion.
46 groups presented their work to mentors from NYU and industry.
The course is a unique opportunity for students to solve real-world problems.
Learn more: cds.nyu.edu/masters-in-d...
46 groups presented their work to mentors from NYU and industry.
The course is a unique opportunity for students to solve real-world problems.
Learn more: cds.nyu.edu/masters-in-d...
January 22, 2026 at 4:27 PM
The CDS Capstone Project course held its Final Poster Presentations at the Rosenthal Pavilion.
46 groups presented their work to mentors from NYU and industry.
The course is a unique opportunity for students to solve real-world problems.
Learn more: cds.nyu.edu/masters-in-d...
46 groups presented their work to mentors from NYU and industry.
The course is a unique opportunity for students to solve real-world problems.
Learn more: cds.nyu.edu/masters-in-d...
CDS PhD student Claudia Skok Gibbs shows that DNA sequence alone can construct powerful priors for gene network prediction — and that in GRN inference, prior quality matters more than which inference algorithm is chosen.
nyudatascience.medium.com/what-dna-alo...
nyudatascience.medium.com/what-dna-alo...
What DNA Alone Can Tell Us About Gene Networks
In GRN inference, performance is bottlenecked not by the inference algorithm, but by the quality of the input prior.
nyudatascience.medium.com
January 21, 2026 at 7:35 PM
CDS PhD student Claudia Skok Gibbs shows that DNA sequence alone can construct powerful priors for gene network prediction — and that in GRN inference, prior quality matters more than which inference algorithm is chosen.
nyudatascience.medium.com/what-dna-alo...
nyudatascience.medium.com/what-dna-alo...
CDS graduate students gathered at Amity Hall Downtown for an end-of-year celebration.
Attendees enjoyed drinks and light refreshments to mark the close of the semester.
Attendees enjoyed drinks and light refreshments to mark the close of the semester.
January 16, 2026 at 7:25 PM
CDS graduate students gathered at Amity Hall Downtown for an end-of-year celebration.
Attendees enjoyed drinks and light refreshments to mark the close of the semester.
Attendees enjoyed drinks and light refreshments to mark the close of the semester.
New research on reward hacking — TRACE detects when reasoning models exploit shortcuts instead of doing the intended work.
From CDS visiting academic Xinpeng Wang, Nitish Joshi, Barbara Plank, CDS Postdoc Researcher Rico Angell, & CDS Assoc. Prof. He He.
nyudatascience.medium.com/catching-the...
From CDS visiting academic Xinpeng Wang, Nitish Joshi, Barbara Plank, CDS Postdoc Researcher Rico Angell, & CDS Assoc. Prof. He He.
nyudatascience.medium.com/catching-the...
Catching the AI That Cheats: New Method Detects Reward Hacking in Reasoning Models
TRACE measures how much reasoning a model needs to answer correctly, helping identify cases where it relies on shortcuts instead.
nyudatascience.medium.com
January 16, 2026 at 4:18 PM
New research on reward hacking — TRACE detects when reasoning models exploit shortcuts instead of doing the intended work.
From CDS visiting academic Xinpeng Wang, Nitish Joshi, Barbara Plank, CDS Postdoc Researcher Rico Angell, & CDS Assoc. Prof. He He.
nyudatascience.medium.com/catching-the...
From CDS visiting academic Xinpeng Wang, Nitish Joshi, Barbara Plank, CDS Postdoc Researcher Rico Angell, & CDS Assoc. Prof. He He.
nyudatascience.medium.com/catching-the...
Attention, MS Applicants! Still working on your application to the CDS MS program? Here’s a friendly reminder of what you need to get over the finish line:
January 15, 2026 at 7:52 PM
Attention, MS Applicants! Still working on your application to the CDS MS program? Here’s a friendly reminder of what you need to get over the finish line:
CDS undergraduates presented original data science projects at the 2025 DSC Project Expo.
Industry judges from CVS Aetna, Amazon, Goguma, and NYC Health + Hospitals evaluated the work.
Industry judges from CVS Aetna, Amazon, Goguma, and NYC Health + Hospitals evaluated the work.
January 14, 2026 at 8:28 PM
CDS undergraduates presented original data science projects at the 2025 DSC Project Expo.
Industry judges from CVS Aetna, Amazon, Goguma, and NYC Health + Hospitals evaluated the work.
Industry judges from CVS Aetna, Amazon, Goguma, and NYC Health + Hospitals evaluated the work.
A single character can reshuffle AI leaderboards.
CDS PhD student Jingtong Su and CDS PhD alum Jianyu Zhang show that changing a delimiter in eval prompts can swing MMLU scores by up to 23 percent, enough to erase years of reported progress.
nyudatascience.medium.com/a-comma-or-a...
CDS PhD student Jingtong Su and CDS PhD alum Jianyu Zhang show that changing a delimiter in eval prompts can swing MMLU scores by up to 23 percent, enough to erase years of reported progress.
nyudatascience.medium.com/a-comma-or-a...
A Comma or a Hashtag? How One Character Can Upend AI Leaderboards
Changing one character in an eval prompt can swing AI benchmark scores enough to upend leaderboard rankings.
nyudatascience.medium.com
January 14, 2026 at 6:40 PM
A single character can reshuffle AI leaderboards.
CDS PhD student Jingtong Su and CDS PhD alum Jianyu Zhang show that changing a delimiter in eval prompts can swing MMLU scores by up to 23 percent, enough to erase years of reported progress.
nyudatascience.medium.com/a-comma-or-a...
CDS PhD student Jingtong Su and CDS PhD alum Jianyu Zhang show that changing a delimiter in eval prompts can swing MMLU scores by up to 23 percent, enough to erase years of reported progress.
nyudatascience.medium.com/a-comma-or-a...
MS students in CDS Assistant Professor of Computer Science and Data Science Eunsol Choi's (@eunsol.bsky.social) "Fundamentals of NLP" course presented final projects in an end-of-year poster session.
Three projects won best poster:
Three projects won best poster:
January 13, 2026 at 4:15 PM
MS students in CDS Assistant Professor of Computer Science and Data Science Eunsol Choi's (@eunsol.bsky.social) "Fundamentals of NLP" course presented final projects in an end-of-year poster session.
Three projects won best poster:
Three projects won best poster:
Thanks to WiDS for organizing a crafty Holiday Chillout before finals.
The event featured DIY snow globe crafting, free pasta, and hot chocolate in the CDS Open Space.
The event featured DIY snow globe crafting, free pasta, and hot chocolate in the CDS Open Space.
January 12, 2026 at 6:38 PM
Thanks to WiDS for organizing a crafty Holiday Chillout before finals.
The event featured DIY snow globe crafting, free pasta, and hot chocolate in the CDS Open Space.
The event featured DIY snow globe crafting, free pasta, and hot chocolate in the CDS Open Space.
CDS alumni gathered at the end of the year for the annual Holiday Party.
President of the CDS Alumni Council Elliot Silva announced the Alumni x CDS Hackathon, set for Saturday, March 28th, with Anand Tyagi as lead organizer.
President of the CDS Alumni Council Elliot Silva announced the Alumni x CDS Hackathon, set for Saturday, March 28th, with Anand Tyagi as lead organizer.
January 9, 2026 at 5:25 PM
CDS alumni gathered at the end of the year for the annual Holiday Party.
President of the CDS Alumni Council Elliot Silva announced the Alumni x CDS Hackathon, set for Saturday, March 28th, with Anand Tyagi as lead organizer.
President of the CDS Alumni Council Elliot Silva announced the Alumni x CDS Hackathon, set for Saturday, March 28th, with Anand Tyagi as lead organizer.
CDS MS alumni report an average salary of $128,000, with recent grads working across tech, finance, education, consulting, and health — from Amazon and Meta to Morgan Stanley and NYU.
nyudatascience.medium.com/by-the-numbe...
nyudatascience.medium.com/by-the-numbe...
By the Numbers: Career Outcomes & Student Success at NYU CDS
CDS shares new career outcomes data, alumni placements, and student perspectives from its MS in Data Science program.
nyudatascience.medium.com
January 9, 2026 at 4:42 PM
CDS MS alumni report an average salary of $128,000, with recent grads working across tech, finance, education, consulting, and health — from Amazon and Meta to Morgan Stanley and NYU.
nyudatascience.medium.com/by-the-numbe...
nyudatascience.medium.com/by-the-numbe...
Principal Researcher Jordan Ash from Microsoft Research recently discussed improving sample efficiency at CDS's CILVR Seminar.
Ash shared VeSSAL, a batch active learning algorithm, and explored how its principles can improve language model efficiency.
Ash shared VeSSAL, a batch active learning algorithm, and explored how its principles can improve language model efficiency.
January 8, 2026 at 9:57 PM
Principal Researcher Jordan Ash from Microsoft Research recently discussed improving sample efficiency at CDS's CILVR Seminar.
Ash shared VeSSAL, a batch active learning algorithm, and explored how its principles can improve language model efficiency.
Ash shared VeSSAL, a batch active learning algorithm, and explored how its principles can improve language model efficiency.
New year, new data science research! Julia Kempe’s new “soft” chain-of-thought training approach, Mengye Ren’s new “ARQ” learning algorithm, and more on this month’s research feature! t.e2ma.net/webview/tpc7...
January 7, 2026 at 8:38 PM
New year, new data science research! Julia Kempe’s new “soft” chain-of-thought training approach, Mengye Ren’s new “ARQ” learning algorithm, and more on this month’s research feature! t.e2ma.net/webview/tpc7...
Courant PhD student @zaynesprague.bsky.social, CDS Associate Professor @gregdnlp.bsky.social, CDS Assistant Professor @mengyer.bsky.social, and others propose a new method for teaching models to reason — even when they don’t start with reasoning skills.
nyudatascience.medium.com/teaching-lan...
nyudatascience.medium.com/teaching-lan...
Teaching Language Models to Think: A New Approach to Learning Reasoning Skills
Durrett and Ren’s SkillFactory method helps models learn cognitive behaviors by reflecting on their own answers—no smarter teacher…
nyudatascience.medium.com
January 7, 2026 at 7:58 PM
Courant PhD student @zaynesprague.bsky.social, CDS Associate Professor @gregdnlp.bsky.social, CDS Assistant Professor @mengyer.bsky.social, and others propose a new method for teaching models to reason — even when they don’t start with reasoning skills.
nyudatascience.medium.com/teaching-lan...
nyudatascience.medium.com/teaching-lan...
The full NYU Agentic AI Workshop video series is now available.
CDS alum Adi Singhal led this four-session workshop covering the Model Context Protocol and how to build intelligent agents.
20+ videos on agentic AI, multi-agent systems, and more.
🔗 www.youtube.com/watch?v=LoTp...
CDS alum Adi Singhal led this four-session workshop covering the Model Context Protocol and how to build intelligent agents.
20+ videos on agentic AI, multi-agent systems, and more.
🔗 www.youtube.com/watch?v=LoTp...
MCP Fundamentals in 20 minutes using the Restaurant Analogy
YouTube video by NYU Data Science
www.youtube.com
January 7, 2026 at 4:43 PM
The full NYU Agentic AI Workshop video series is now available.
CDS alum Adi Singhal led this four-session workshop covering the Model Context Protocol and how to build intelligent agents.
20+ videos on agentic AI, multi-agent systems, and more.
🔗 www.youtube.com/watch?v=LoTp...
CDS alum Adi Singhal led this four-session workshop covering the Model Context Protocol and how to build intelligent agents.
20+ videos on agentic AI, multi-agent systems, and more.
🔗 www.youtube.com/watch?v=LoTp...
CDS Associate Professor Brian McFee shares essential skills for data science grads in conversation with MS Admissions Ambassador Vidhi Manek.
January 5, 2026 at 5:00 PM
CDS Associate Professor Brian McFee shares essential skills for data science grads in conversation with MS Admissions Ambassador Vidhi Manek.