Tengyu Ma 0001 - researchr alias

researchr

You are not signed in
Sign in
Sign up

Viewing Publication 1 - 37 from 37

2023

What learning algorithm is in-context learning? Investigations with linear modelsEkin Akyürek, Dale Schuurmans, Jacob Andreas, Tengyu Ma 0001, Denny Zhou. iclr 2023: [doi]

Asymptotic Instance-Optimal Algorithms for Interactive Decision MakingKefan Dong, Tengyu Ma 0001. iclr 2023: [doi]

First Steps Toward Understanding the Extrapolation of Nonlinear Models to Unseen DomainsKefan Dong, Tengyu Ma 0001. iclr 2023: [doi]

Toward L_∞Recovery of Nonlinear Functions: A Polynomial Sample Complexity Bound for Gaussian Random FieldsKefan Dong, Tengyu Ma 0001. colt 2023: 2877-2918 [doi]

What is the Inductive Bias of Flatness Regularization? A Study of Deep Matrix Factorization ModelsKhashayar Gatmiry, Zhiyuan Li 0001, Tengyu Ma 0001, Sashank Reddi, Stefanie Jegelka, Ching-Yao Chuang. nips 2023: [doi]

Max-Margin Works while Large Margin Fails: Generalization without Uniform ConvergenceMargalit Glasgow, Colin Wei, Mary Wootters, Tengyu Ma 0001. iclr 2023: [doi]

A theoretical study of inductive biases in contrastive learningJeff Z. HaoChen, Tengyu Ma 0001. iclr 2023: [doi]

Same Pre-training Loss, Better Downstream: Implicit Bias Matters for Language ModelsHong Liu, Sang Michael Xie, Zhiyuan Li, Tengyu Ma 0001. icml 2023: 22188-22214 [doi]

Beyond NTK with Vanilla Gradient Descent: A Mean-Field Analysis of Neural Networks with Polynomial Width, Samples, and TimeArvind Mahankali, Haochen Zhang, Kefan Dong, Margalit Glasgow, Tengyu Ma 0001. nips 2023: [doi]

Symbol tuning improves in-context learning in language modelsJerry W. Wei, Le Hou, Andrew K. Lampinen, Xiangning Chen, Da Huang, Yi Tay, Xinyun Chen, Yifeng Lu, Denny Zhou, Tengyu Ma 0001, Quoc V. Le. emnlp 2023: 968-979 [doi]

Sharpness Minimization Algorithms Do Not Only Minimize Sharpness To Achieve Better GeneralizationKaiyue Wen, Zhiyuan Li 0005, Tengyu Ma 0001. nips 2023: [doi]

How Sharpness-Aware Minimization Minimizes Sharpness?Kaiyue Wen, Tengyu Ma 0001, Zhiyuan Li. iclr 2023: [doi]

Data Selection for Language Models via Importance ResamplingSang Michael Xie, Shibani Santurkar, Tengyu Ma 0001, Percy Liang. nips 2023: [doi]

DoReMi: Optimizing Data Mixtures Speeds Up Language Model PretrainingSang Michael Xie, Hieu Pham 0001, Xuanyi Dong, Nan Du, Hanxiao Liu, Yifeng Lu, Percy Liang, Quoc V. Le, Tengyu Ma 0001, Adams Wei Yu. nips 2023: [doi]

2022

On the optimization landscape of tensor decompositionsRong Ge 0001, Tengyu Ma 0001. mp, 193(2):713-759, 2022. [doi]

Near-Optimal Algorithms for Autonomous Exploration and Multi-Goal Stochastic Shortest PathHaoyuan Cai, Tengyu Ma 0001, Simon S. Du. icml 2022: 2434-2456 [doi]

Iterative Feature Matching: Toward Provable Domain Generalization with Logarithmic EnvironmentsYining Chen, Elan Rosenfeld, Mark Sellke, Tengyu Ma 0001, Andrej Risteski. nips 2022: [doi]

Sharp Bounds for Federated Averaging (Local SGD) and Continuous PerspectiveMargalit R. Glasgow, Honglin Yuan, Tengyu Ma 0001. aistats 2022: 9050-9090 [doi]

Beyond Separability: Analyzing the Linear Transferability of Contrastive Representations to Related SubpopulationsJeff Z. HaoChen, Colin Wei, Ananya Kumar, Tengyu Ma 0001. nips 2022: [doi]

Calibrated ensembles can mitigate accuracy tradeoffs under distribution shiftAnanya Kumar, Tengyu Ma 0001, Percy Liang, Aditi Raghunathan. uai 2022: 1041-1051 [doi]

Fine-Tuning can Distort Pretrained Features and Underperform Out-of-DistributionAnanya Kumar, Aditi Raghunathan, Robbie Matthew Jones, Tengyu Ma 0001, Percy Liang. iclr 2022: [doi]

DR3: Value-Based Deep Reinforcement Learning Requires Explicit RegularizationAviral Kumar, Rishabh Agarwal, Tengyu Ma 0001, Aaron C. Courville, George Tucker, Sergey Levine. iclr 2022: [doi]

Self-supervised Learning is More Robust to Dataset ImbalanceHong Liu, Jeff Z. HaoChen, Adrien Gaidon, Tengyu Ma 0001. iclr 2022: [doi]

Plan Better Amid Conservatism: Offline Multi-Agent Reinforcement Learning with Actor RectificationLing Pan, Longbo Huang, Tengyu Ma 0001, Huazhe Xu. icml 2022: 17221-17237 [doi]

Connect, Not Collapse: Explaining Contrastive Learning for Unsupervised Domain AdaptationKendrick Shen, Robbie M. Jones, Ananya Kumar, Sang Michael Xie, Jeff Z. HaoChen, Tengyu Ma 0001, Percy Liang. icml 2022: 19847-19878 [doi]

Statistically Meaningful Approximation: a Case Study on Approximating Turing Machines with TransformersColin Wei, Yining Chen, Tengyu Ma 0001. nips 2022: [doi]

An Explanation of In-context Learning as Implicit Bayesian InferenceSang Michael Xie, Aditi Raghunathan, Percy Liang, Tengyu Ma 0001. iclr 2022: [doi]

2021

Entity and Evidence Guided Document-Level Relation ExtractionKevin Huang 0002, Peng Qi 0003, Guangtao Wang, Tengyu Ma 0001, Jing Huang 0019. rep4nlp 2021: 307-315 [doi]

Label Noise SGD Provably Prefers Flat Global MinimizersAlex Damian, Tengyu Ma 0001, Jason D. Lee. nips 2021: 27449-27461 [doi]

Provable Model-based Nonlinear Bandit and Reinforcement Learning: Shelve Optimism, Embrace Virtual CurvatureKefan Dong, Jiaqi Yang, Tengyu Ma 0001. nips 2021: 26168-26182 [doi]

Shape Matters: Understanding the Implicit Bias of the Noise CovarianceJeff Z. HaoChen, Colin Wei, Jason D. Lee, Tengyu Ma 0001. colt 2021: 2315-2357 [doi]

Provable Guarantees for Self-Supervised Deep Learning with Spectral Contrastive LossJeff Z. HaoChen, Colin Wei, Adrien Gaidon, Tengyu Ma 0001. nips 2021: 5000-5011 [doi]

Learning Barrier Certificates: Towards Safe Reinforcement Learning with Zero Training-time ViolationsYuping Luo, Tengyu Ma 0001. nips 2021: 25621-25632 [doi]

Safe Reinforcement Learning by Imagining the Near FutureGarrett Thomas, Yuping Luo, Tengyu Ma 0001. nips 2021: 13859-13869 [doi]

Why Do Pretrained Language Models Help in Downstream Tasks? An Analysis of Head and Prompt TuningColin Wei, Sang Michael Xie, Tengyu Ma 0001. nips 2021: 16158-16170 [doi]

Fine-Grained Gap-Dependent Bounds for Tabular MDPs via Adaptive Multi-Step BootstrapHaike Xu, Tengyu Ma 0001, Simon S. Du. colt 2021: 4438-4472 [doi]

Calibrating Predictions to Decisions: A Novel Approach to Multi-Class CalibrationShengjia Zhao, Michael P. Kim, Roshni Sahoo, Tengyu Ma 0001, Stefano Ermon. nips 2021: 22313-22324 [doi]

See Also

All Publications

Filter by Year
OR AND NOT 1

Filter by Type
OR AND NOT 1

Filter by Tag

Filter by Venue
OR AND NOT 1

Filter by Author

[+]
OR AND NOT 1

Filter by Top Terms

[+]
OR AND NOT 1

Researchr

Publications by 'Tengyu Ma 0001'

Viewing Publication 1 - 37 from 37

2023

2022

2021

See Also

All Publications

Filter by YearOR AND NOT 1

Filter by TypeOR AND NOT 1

Filter by Tag

Filter by VenueOR AND NOT 1

Filter by Author [+]OR AND NOT 1

Filter by Top Terms [+]OR AND NOT 1

Researchr

Publications by 'Tengyu Ma 0001'

Viewing Publication 1 - 37 from 37

2023

2022

2021

Filter by Year
OR AND NOT 1

Filter by Type
OR AND NOT 1

Filter by Venue
OR AND NOT 1

Filter by Author

[+]
OR AND NOT 1

Filter by Top Terms

[+]
OR AND NOT 1