Greetings

I am a postdoc at the Simons Institute for the Theory of Computing at UC Berkeley, where I work with Peter Bartlett and Jason D. Lee. In 2027, I will be joining the faculty in the School of Computer Science at the University of Sydney.

I completed my PhD at the University of Illinois, Urbana-Champaign, where I was advised by the eminent Nan Jiang. During my PhD, I’ve also been fortunate to work with magnificent researchers Dylan J. Foster and Akshay Krishnamurty (Microsoft Research), Dean P. Foster (Amazon Research), and Csaba Szepesvári (University of Alberta). Before that, in a distant land, I completed a MSc in Computer Science (advised by the formidable Prakash Panangaden and Marc G. Bellemare) and a BSc in Maths & Physics at McGill University.

Here are links to my Google Scholar and my perpetually outdated CV. I can be reached at p.amortila@berkeley.edu.

Education

  • 2019 — 2025. PhD in Computer Science at University of Illinois, Urbana-Champaign.

  • 2017 — 2019. MSc in Computer Science at McGill University.

  • 2013 — 2017. BSc in Honours Maths & Physics at McGill University.

    • Distinctions: First Class Honours, Principal’s Student-Athlete Honour Roll.

Positions

Selected Publications

  1. When Does On-Policy Interaction Help? Representational Tradeoffs in Value-Based Imitation Learning [arXiv]

    Luca Viano, Antoine Moulin, Audrey Huang, Volkan Cevher, Philip Amortila, Dylan J. Foster

    NeurIPS 2026

    European Workshop On RL (EWRL) 2026 Oral Presentation

  2. Reinforcement Learning under Latent Dynamics: Toward Statistical and Algorithmic Modularity [arXiv, talk]

    Philip Amortila, Dylan J. Foster, Nan Jiang, Akshay Krishnamurthy, Zakaria Mhammedi

    NeurIPS 2024 Oral Presentation

  3. Scalable Online Exploration via Coverability [arXiv, talk]

    Philip Amortila, Dylan J. Foster, Akshay Krishnamurthy

    ICML 2024

  4. Exponential Lower Bounds for Planning in MDPs With Linearly-Realizable Optimal Action-Value Functions [arXiv, talk]

    Gellert Weisz, Philip Amortila, Csaba Szepesvári

    ALT 2021 Best Student Paper Award

All Publications

Preprints

  1. Post-Training at the Edge of Detectability: A Game-Theoretic Approach to Fine-Tuning[arXiv]

    Keegan Harris, Brian W. Lee, Ian Waudby-Smith, Philip Amortila, Nika Haghtalab, Michael I. Jordan

    Preprint

Conference Papers

  1. When Does On-Policy Interaction Help? Representational Tradeoffs in Value-Based Imitation Learning [arXiv]

    Luca Viano, Antoine Moulin, Audrey Huang, Volkan Cevher, Philip Amortila, Dylan J. Foster

    NeurIPS 2026

    European Workshop On RL (EWRL) 2026 Oral Presentation

  2. A Unifying View of Coverage in Linear Off-Policy Evaluation [arXiv]

    Philip Amortila, Audrey Huang, Akshay Krishnamurthy, Nan Jiang

    ICLR 2026

  3. Model Selection for Off-Policy Evaluation: New Algorithms and Experimental Protocol [arXiv]

    Pai Liu, Lingfeng Zhao, Shivangi Agarwal, Jinghan Liu, Audrey Huang, Philip Amortila, Nan Jiang

    NeurIPS 2025

  4. Reinforcement Learning under Latent Dynamics: Toward Statistical and Algorithmic Modularity [arXiv, talk]

    Philip Amortila, Dylan J. Foster, Nan Jiang, Akshay Krishnamurthy, Zakaria Mhammedi

    NeurIPS 2024 Oral Presentation

  5. Mitigating Covariate Shift in Misspecified Regression with Applications to Reinforcement Learning [arXiv]

    Philip Amortila, Tongyi Cao, Akshay Krishnamurthy

    COLT 2024

  6. Scalable Online Exploration via Coverability [arXiv, talk]

    Philip Amortila, Dylan J. Foster, Akshay Krishnamurthy

    ICML 2024

  7. Harnessing Density Ratios for Online Reinforcement Learning [arXiv]

    Philip Amortila, Dylan J. Foster, Nan Jiang, Ayush Sekhari, Tengyang Xie

    ICLR 2024 Spotlight Presentation

  8. The Optimal Approximation Factors in Misspecified Off-Policy Value Function Estimation [arXiv]

    Philip Amortila, Nan Jiang, Csaba Szepesvari

    ICML 2023

  9. A Few Expert Queries Suffices for Sample-Efficient RL with Resets and Linear Value Approximation [arXiv, talk]

    Philip Amortila, Nan Jiang, Dhruv Madeka, Dean P. Foster

    NeurIPS 2022

  10. On Query-efficient Planning in MDPs under Linear Realizability of the Optimal State-value Function [arXiv, talk]

    Gellert Weisz, Philip Amortila, Barnabàs Janzer, Yasin Abbasi-Yadkori, Nan Jiang, Csaba Szepesvári 

    COLT 2021

  11. Exponential Lower Bounds for Planning in MDPs With Linearly-Realizable Optimal Action-Value Functions [arXiv, talk]

    Gellert Weisz, Philip Amortila, Csaba Szepesvári

    ALT 2021 Best Student Paper Award

  12. Solving Constrained Markov Decision Processes via Backward Value Functions [arXiv, talk]

    Harsh Satija, Philip Amortila, Joelle Pineau

    ICML 2020

  13. A Distributional Analysis of Sampling-Based Reinforcement Learning Algorithms [arXiv, talk]

    Philip Amortila, Doina Precup, Prakash Panangaden, Marc G. Bellemare

    AISTATS 2020

    NeurIPS 2019 Optimization in RL Workshop Spotlight Presentation [talk]

  14. Learning Graph Weighted Models on Pictures [arXiv]

    Philip Amortila, Guillaume Rabusseau

    ICGI 2018

Workshop Papers

  1. Temporally Extended Metrics for Markov Decision Processes [pdf]

    Philip Amortila, Marc G. Bellemare, Prakash Panangaden, Doina Precup

    AAAI 2019 Safety in AI Workshop Spotlight

Technical Notes

  1. A Variant of the Wang-Foster-Kakade Lower Bound for the Discounted Setting [arXiv]

    Philip Amortila, Nan Jiang, Tengyang Xie

    Technical Note

  • 2023. Finalist for Google PhD Fellowship (2023).

    • Nominated by UIUC for national competition (3 selected among all UIUC students).

  • 2022. Finalist for Apple PhD Fellowship (2022)

    • Nominated by UIUC for national competition (3 selected among all UIUC students).

  • 2021. Best Student Paper Award at ALT 2021.

  • 2019. NSERC Postgraduate Doctoral Fellowship (PGS-D).

Selected Awards