Eric Zhao

Eric Zhao

I'm a 4th year PhD candidate in CS at UC Berkeley, where I'm advised by Michael I. Jordan and Nika Haghtalab. I'm a Google PhD Fellow, a NSF Graduate Research Fellow, and affiliated with the Berkeley AI Research Lab (BAIR).

My current research focuses on enabling language models to reason in low-resource domains where verification is difficult, such as autonomously proving results in niche theoretical topics. I also enjoy manually proving things; in particular, my theoretical research studies the alignment and truthfulness challenges inherent in building multi-objective forecasting and learning systems.

I received my B.S. from Caltech in 2020 and previously interned with Google Research, Nvidia Research, and Salesforce Research.

Selected Works

(α-β) denotes when authors are ordered alphabetically.

  • Sample, Scrutinize and Scale: Effective Inference-Time Search by Scaling Verification Eric Zhao, Pranjal Awasthi, Sreenivas Gollapudi
    PDF Blog Preprint, Jan 2025
  • Truthfulness of Decision-Theoretic Calibration Measures (α-β) Mingda Qiao, Eric Zhao
    PDF Blog Preprint, Jan 2025
  • From Style to Facts: Mapping the Boundaries of Knowledge Injection with Finetuning Eric Zhao, Pranjal Awasthi, Nika Haghtalab
    PDF Blog Preprint, Feb 2025
  • On-Demand Sampling: Learning Optimally from Multiple Distributions (α-β) Nika Haghtalab, Michael I. Jordan, Eric Zhao
    PDF Blog NeurIPS 2022
    Neurips Outstanding Paper Award

Selected Awards

  • Google PhD Fellowship (2024)
  • NSF Graduate Research Fellowship (2023)
  • NeurIPS Best Paper Award (2022)

Contact

eric.zh@berkeley.edu