Gregory Valiant

Timeline

1. 2006–2012

   Degrees at Harvard and Berkeley

   His degrees from Harvard University include a BA in mathematics, awarded in 2006, and an MS in computer science. In 2012, he earns a computer science PhD in UC Berkeley’s EECS Department, advised by Christos Papadimitriou, with the dissertation Algorithmic Approaches to Statistical Questions. The dissertation receives an honorable mention for the 2012 ACM Doctoral Dissertation Award.

2. 2013

   Joining Stanford computer science

   In 2013, he joins Stanford’s Computer Science Department as an assistant professor after a postdoctoral post at Microsoft Research New England.

3. 2014–2016

   CAREER and Sloan awards

   He receives an NSF CAREER Award in 2014 and a Sloan Foundation Fellowship in 2016.

4. 2016–2017

   Learning from untrusted data

   With Moses Charikar and Jacob Steinhardt, he coauthors Learning from Untrusted Data, first posted in 2016 and revised in 2017. The paper studies learning from biased or malicious data through two frameworks: producing a list that contains an accurate answer, and using a small trusted dataset to extract information from a larger untrusted dataset.

5. 2019

   Making AI forget you

   In 2019, he coauthors Making AI Forget You: Data Deletion in Machine Learning with Antonio Ginart, Melody Guan and James Zou at NeurIPS. The work formalizes the problem of deleting individual data points from trained models and proposes two efficient deletion algorithms for k-means clustering.

6. 2020

   Sample amplification

   With Brian Axelrod, Shivam Garg and Vatsal Sharan, he publishes Sample Amplification: Increasing Dataset Size even when Learning is Impossible at ICML in 2020. The work gives procedures that enlarge datasets while preserving the appearance of independent samples, for discrete distributions and Gaussian distributions with unknown means and fixed covariance.

7. 2022

   Transformers learn from prompts

   In 2022, he coauthors What Can Transformers Learn In-Context? A Case Study of Simple Function Classes with Shivam Garg, Dimitris Tsipras and Percy Liang. Their experiments show that transformers trained from scratch can learn unseen linear functions from examples in a prompt, with performance comparable to an optimal least squares estimator.

8. 2025

   Equivalent statistical adversaries

   With Guy Blanc, he publishes Adaptive and Oblivious Statistical Adversaries Are Equivalent at STOC in 2025. They prove that adversaries with access to the sampled data and adversaries that choose corruption without that access have equivalent power up to polynomial factors in sample size.

9. As of October 2026

   Associate professor at Stanford

   As of October 2026, he is an associate professor of computer science at Stanford University.