NeurIPS 2025 presentations from our group members

Check the NeurIPS 2025 Presentations from our group!

Main Meeting

1. Don't be lazy: CompleteP enables compute-efficient deep transformers
Nolan Dey · Bin Zhang · Lorenzo Noci · Mufan Li · Blake Bordelon · Shane Bergsma · Cengiz Pehlevan · Boris Hanin · Joel Hestness
Exhibit Hall C,D,E #4000
Wed 3 Dec 4:30 p.m. PST — 7:30 p.m. PST

2. An Analytical Theory of Spectral Bias in the Learning Dynamics of Diffusion Models (Spotlight)
Binxu Wang · Cengiz Pehlevan
Exhibit Hall C,D,E #2904
Fri 5 Dec 4:30 p.m. PST — 7:30 p.m. PST

3. A solvable model of learning generative diffusion: theory and insights
Hugo Cui · Cengiz Pehlevan · Yue Lu
Exhibit Hall C,D,E #3011
Thu 4 Dec 11 a.m. PST — 2 p.m. PST

4. Error Broadcast and Decorrelation as a Potential Artificial and Natural Learning Mechanism (Spotlight)
Mete Erdogan · Cengiz Pehlevan · Alper Erdogan
Exhibit Hall C,D,E #2006
Wed 3 Dec 11 a.m. PST — 2 p.m. PST

Workshops

5. Perspective: Summary Statistics of Learning
Jacob A. Zavatone-Veth · Blake Bordelon · Cengiz Pehlevan
UniReps 

6. Pretrain-Test Task Alignment Model for In-Context Learning by Linear Attention
Mary I. Letey · Yue M. Lu · Cengiz Pehlevan
What Can(’t) Transformers Do?

7. A Simplified Analysis of SGD for Linear Regression with Weight Averaging
Alexandru Meterez · Depen Morwani · Costin-Andrei Oncescu · Jingfeng Wu · Cengiz Pehlevan · Sham Kakade
OPT 2025

8. How does training shape the Riemannian geometry of neural network representations? (Oral)
 Jacob A. Zavatone-Veth · Sheng Yang · Julian A. Rubinfien · Cengiz Pehlevan
NeurReps 2025

9. Persistent Homology Distances for Comparing Disease-Filtered Structural Connectomes
Hamza T. Chaudhry · Vineet R. Tiruvadi · Cengiz Pehlevan · Michael D. Fox · Kanaka Rajan
NeurReps 2025

10. Is the Brain Mechanistically Identifiable? (Invited Talk)
Cengiz Pehlevan
Data on the Brain & Mind