CP
Cengiz Pehlevan
cs.LGstat.MLcond-mat.dis-nnq-bio.NCcs.AIcs.NEneurosciencehep-thNeural Networks, Computercs.CV
On Valency
published · living versionsW_3r3u6gpt·v1 · currentpublished
Weighted State Souping for Efficient Long-Context Retrieval
with William L. Tong, Aryo Lotfi, Emmanuel Abbe, Josh Susskind +1
1 version
Preprints & journals
142 papers in the corpus · 2007–2026Weighted State Souping for Efficient Long-Context Retrieval2610.25621v1 · William L. Tong, Aryo Lotfi, Emmanuel Abbe et al.2026 · 0 citationsarXivon Valency
In-context classification: Calibration and noise in linear attention with softmax readout2610.17061v1 · Arman Rysmakhanov, Mary Letey, Yue M. Lu et al.2026 · 0 citationsarXiv
A network model of path control.42926497 · Osman, Mohammed Abdal Monium, Zhang, Wenyi, Pehlevan, Cengiz et al.2026 · 1 citationbioRxiv : the preprint server for biology. 2026
What can linear attention learn from nonlinear teachers in-context?2610.10761v1 · Mary Letey, Arman Rysmakhanov, Yue M. Lu et al.2026 · 0 citationsarXiv
A Solvable Model of Adaptive Learning Rate Rescaling: Acceleration, Stability & Scaling2610.06701v1 · Itay Lavie, Clarissa Lauditi, Cengiz Pehlevan2026 · 0 citationsarXiv
Efficiency and robustness partition the solution space for memory in recurrent networks2610.04697v1 · William Qian, Cengiz Pehlevan2026 · 0 citationsarXiv
A Solvable Theory of Pre-training Data Poisoning: Regime-Dependent Scaling Exponents2609.32288v1 · Indranil Halder, Rastri Dey, Cengiz Pehlevan2026 · 0 citationsarXiv
On the Importance of Gating: Memorization vs. In-Context Learning in State Space Models2609.16540v1 · William L. Tong, Aryo Lotfi, Emmanuel Abbe et al.2026 · 0 citationsarXiv
A Defense of the Quadratic Model2607.21716v1 · Alexandru Meterez, Pranav Ajit Nair, Depen Morwani et al.2026 · 0 citationsarXiv
Sequential Correlations Change In-Context Learning: Effective Context Length and Architectural Mismatch2607.03660v1 · Mary Letey, Yue M. Lu, Cengiz Pehlevan et al.2026 · 0 citationsarXiv
Score Broadcast and Decorrelation: A General Framework for Broadcast-Based Credit Assignment2605.30638v2 · Mustafa Uzun, Mete Erdogan, Cengiz Pehlevan et al.2026 · 0 citationsarXiv
Structure, disorder, and dynamics in task-trained recurrent neural circuits.42239479 · Clark, David G, Bordelon, Blake, Zavatone-Veth, Jacob A et al.2026 · 2 citationsbioRxiv : the preprint server for biology. 2026
Spectral Dynamics in Deep Networks: Feature Learning, Outlier Escape, and Learning Rate Transfer2605.07870v2 · Clarissa Lauditi, Cengiz Pehlevan, Blake Bordelon2026 · 0 citationsarXiv
Author Correction: Predictive coding of reward in the hippocampus.42120502 · Yaghoubi, Mohammad, Kumar, M Ganesh, Nieto-Posadas, Andres et al.2026 · 0 citationsNature. 2026;653(8116):E16
Jailbreak Scaling Laws for Large Language Models: Polynomial-Exponential Crossover2603.11331v4 · Indranil Halder, Annesya Banerjee, Cengiz Pehlevan2026 · 0 citationsarXiv
Boule or Baguette? A Study on Task Topology, Length Generalization, and the Benefit of Reasoning Traces2602.14404v1 · William L. Tong, Ege Cakar, Cengiz Pehlevan2026 · 0 citationsarXiv
Universal One-third Time Scaling in Learning Peaked Distributions2602.03685v3 · Yizhou Liu, Ziming Liu, Cengiz Pehlevan et al.2026 · 0 citationsarXiv
Anytime Pretraining: Horizon-Free Learning-Rate Schedules with Weight Averaging2602.03702v2 · Alexandru Meterez, Pranav Ajit Nair, Depen Morwani et al.2026 · 0 citationsarXiv
A Random Matrix Theory Perspective on the Consistency of Diffusion Models2602.02908v2 · Binxu Wang, Jacob Zavatone-Veth, Cengiz Pehlevan2026 · 0 citationsarXiv
Hyperparameter Transfer with Mixture-of-Expert Layers2601.20205v3 · Tianze Jiang, Blake Bordelon, Cengiz Pehlevan et al.2026 · 0 citationsarXiv
A note on the dynamics of extended-context disordered kinetic spin models.41583832 · Zavatone-Veth, Jacob A, Pehlevan, Cengiz2026 · 0 citationsJournal of physics. A, Mathematical and theoretical. 2026;59(4):045001
Predictive coding of reward in the hippocampus.41535460 · Yaghoubi, Mohammad, Kumar, M Ganesh, Nieto-Posadas, Andres et al.2026 · 13 citationsNature. 2026;651(8105):414-420
Disordered Dynamics in High Dimensions: Connections to Random Matrices and Machine Learning2601.01010v2 · Blake Bordelon, Cengiz Pehlevan2026 · 0 citationsarXiv
Demystifying LLM-as-a-Judge: Analytically Tractable Model for Inference-Time Scaling2512.19905v3 · Indranil Halder, Cengiz Pehlevan2025 · 0 citationsarXiv
Simultaneous detection and estimation in olfactory sensing.41279341 · Jiang, Chen, He, Matthew Y, Murthy, Venkatesh N et al.2025 · 0 citationsbioRxiv : the preprint server for biology. 2025
Seesaw: Accelerating Training by Balancing Learning Rate and Batch Size Scheduling2510.14717v2 · Alexandru Meterez, Depen Morwani, Jingfeng Wu et al.2025 · 0 citationsarXiv
Theory of Scaling Laws for In-Context Regression: Depth, Width, Context and Time2510.01098v1 · Blake Bordelon, Mary I. Letey, Cengiz Pehlevan2025 · 0 citationsarXiv
Pretrain-Test Task Alignment Governs Generalization in In-Context Learning2509.26551v1 · Mary I. Letey, Jacob A. Zavatone-Veth, Yue M. Lu et al.2025 · 0 citationsarXiv
Discovering alternative solutions beyond the simplicity bias in recurrent neural networks2509.21504v1 · William Qian, Cengiz Pehlevan2025 · 0 citationsarXiv
Summary statistics of learning link changing neural representations to behavior.40313661 · Zavatone-Veth, Jacob A, Bordelon, Blake, Pehlevan, Cengiz2025 · 1 citationArXiv. 2025
Asymptotic theory of in-context learning by linear attention.40632569 · Lu, Yue M, Letey, Mary, Zavatone-Veth, Jacob A et al.2026 · 10 citationsProceedings of the National Academy of Sciences of the United States of America. 2025;122(28):e2502599122
Transfer Learning in Infinite Width Feature Learning Networks2507.04448v2 · Clarissa Lauditi, Blake Bordelon, Cengiz Pehlevan2025 · 0 citationsarXiv
A Simplified Analysis of SGD for Linear Regression with Weight Averaging2506.15535v1 · Alexandru Meterez, Depen Morwani, Costin-Andrei Oncescu et al.2025 · 0 citationsarXiv
Don't be lazy: CompleteP enables compute-efficient deep transformers2505.01618v4 · Nolan Dey, Bin Claire Zhang, Lorenzo Noci et al.2025 · 1 citationarXiv
Error Broadcast and Decorrelation as a Potential Artificial and Natural Learning Mechanism2504.11558v3 · Mete Erdogan, Cengiz Pehlevan, Alper T. Erdogan2025 · 0 citationsarXiv
Echo Chamber: RL Post-training Amplifies Behaviors Learned in Pretraining2504.07912v2 · Rosie Zhao, Alexandru Meterez, Sham Kakade et al.2025 · 0 citationsarXiv
Dynamics of learning to integrate in linear recurrent neural networks2503.18754v2 · Blake Bordelon, Jordan Cotler, Cengiz Pehlevan et al.2025 · 1 citationarXiv
Dynamically Learning to Integrate in Recurrent Neural Networks.40196143 · Bordelon, Blake, Cotler, Jordan, Pehlevan, Cengiz et al.2025 · 2 citationsArXiv. 2025
Neurocomputational underpinnings of suboptimal beliefs in recurrent neural network-based agents10.1101/2025.03.13.642273v1 · Kumar, M. G., Manoogian, A., Qian, W. et al.2025 · 0 citationsbioRxiv
Learning richness modulates equality reasoning in neural networks2503.09781v3 · William L. Tong, Cengiz Pehlevan2025 · 0 citationsarXiv
An Analytical Theory of Spectral Bias in the Learning Dynamics of Diffusion Models2503.03206v3 · Binxu Wang, Cengiz Pehlevan2025 · 0 citationsarXiv
Adaptive kernel predictors from feature-learning infinite limits of neural networks2502.07998v2 · Clarissa Lauditi, Blake Bordelon, Cengiz Pehlevan2025 · 1 citationarXiv
Two-Point Deterministic Equivalence for Stochastic Gradient Dynamics in Linear Models2502.05074v3 · Alexander Atanasov, Blake Bordelon, Jacob A. Zavatone-Veth et al.2025 · 1 citationAdvances in Theoretical and Mathematical Physics (2026) 30, 1
Deep Linear Network Training Dynamics from Random Initialization: Data, Width, Depth, and Hyperparameter Transfer2502.02531v3 · Blake Bordelon, Cengiz Pehlevan2025 · 0 citationsarXiv
A solvable model of learning generative diffusion: theory and insights2501.03937v2 · Hugo Cui, Cengiz Pehlevan, Yue M. Lu2025 · 1 citationarXiv
A Model of Place Field Reorganization During Reward Maximization10.1101/2024.12.12.627755v1 · Kumar, M. G., Bordelon, B., Zavatone-Veth, J. A. et al.2024 · 3 citationsbioRxiv
No Free Lunch From Random Feature Ensembles: Scaling Laws and Near-Optimality Conditions2412.05418v2 · Benjamin S. Ruben, William L. Tong, Hamza Tahir Chaudhry et al.2024 · 0 citationsarXiv
Scaling Laws for Precision2411.04330v2 · Tanishq Kumar, Zachary Ankner, Benjamin F. Spector et al.2024 · 2 citationsarXiv
Do Mice Grok? Glimpses of Hidden Progress During Overtraining in Sensory Cortex2411.03541v2 · Tanishq Kumar, Blake Bordelon, Cengiz Pehlevan et al.2024 · 1 citationarXiv
Bilateral Alignment of Receptive Fields in the Olfactory Cortex.39433407 · Grimaud, Julien, Dorrell, William, Jayakumar, Siddharth et al.2024 · 10 citationseNeuro. 2024;11(11)
Career total: 189 works. 142 are in this corpus.Showing the 50 most recent.
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in alpha. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.