Sara Fish

I am a fourth-year PhD student at Harvard advised by Yannai Gonczarowski. My research interests lie in EconCS and ML. I am supported by a NSF Graduate Research Fellowship and a Kempner Institute Graduate Fellowship. Previously, I completed my B.S. in Mathematics at Caltech.

I am always happy to chat with new people. Feel free to email me at sfish@g.harvard.edu or schedule a meeting here.

Research

(Google Scholar) (arXiv)

EconEvals: Benchmarks and Litmus Tests for Economic Decision-Making by LLM Agents
Sara Fish, Julia Shephard, Minkai Li, Ran Shorrer, and Yannai Gonczarowski
Preprint available at arXiv:2503.18825. (Code)

Generative Social Choice: The Next Generation
with Niclas Boehmer and Ariel Procaccia
ICML 2025 Oral (Top 1%) (arXiv:2505.22939). (Code)

Algorithmic Collusion by Large Language Models
with Yannai Gonczarowski and Ran Shorrer
Preprint available at arXiv:2404.00806.
Coverage: Marginal Revolution, Zvi Mowshowitz, The Capitol Forum, CPI TechREG Chronicle

Stable Menus of Public Goods: A Matching Problem
with Yannai Gonczarowski and Sergiu Hart
EC 2025 (arXiv:2402.11370). (Code)

Generative Social Choice
with Paul Gölz, David Parkes, Ariel Procaccia, Gili Rusak, Itai Shapira, and Manuel Wüthrich
Journal of the ACM (2026) (open access PDF), previously EC 2024. (Code)
(Project awarded $100,000 grant via OpenAI's Democratic Inputs to AI program.)

Crowd Science

I enjoy making small contributions to large scientific endeavors. Listed below are "crowd science" papers I coauthored as a middle (i.e., minor) author.

Prosocial Ranking Challenge (blog post, paper forthcoming)
This is a forthcoming study examining how interventions on social media ranking algorithms impact affective polarization among users. Our team's ranking algorithm submission was selected to be tested (top 3 of 21).

Humanity's Last Exam (arXiv:2501.14249)
This is a LLM benchmark consisting of questions submitted by researchers. Three questions I submitted were included, including one prize-winning question (top 550).

Comparing Human-Only, AI-Assisted, and AI-Led Teams on Assessing Research Reproducibility in Quantitative Social Science (EconPapers URL)
This study evaluates the effectiveness of AI at assisting or conducting reproducibility checks of economics papers. I contributed to the data-gathering process.

Teaching

(Harvard) CS 37 / ECON 1071 ("Incentives in the Wild"): Teaching Fellow, Spring '25
(Harvard) CS 37 / ECON 1071 ("Incentives in the Wild"): Teaching Fellow, Fall '23
(Caltech) Ma 0 ("Intro to Proofs"): Lead Teaching Assistant, Summer '21
(Caltech) Ma 6a ("Discrete Mathematics"): Teaching Assistant, Fall '19
(Caltech) Ma 108a ("Real Analysis"): Grader, Fall '19

Previous volunteering and employment

In summer 2025, I was a Research Intern at Microsoft Research New England.
In summer 2023, I was a mentor for the Polymath Jr. distinct distances project.
In summer 2022, I worked as an Algo Developer Intern at Hudson River Trading.
In summer 2021, I was a counselor at PROMYS, head TA for Caltech's Math 0, and a mentor at the Polymath Jr. expanding polynomials project.
I volunteered for CHMMC (Caltech Harvey Mudd Math Competition) and wrote the 2019 Power Round.
In summer 2018, I worked as a JC at Canada/USA Mathcamp.

Previous math research

A Construction for Difference Sets with Local Properties
with Ben Lund and Adam Sheffer
European Journal of Combinatorics, Vol. 79, June 2019, pg. 237-243. arXiv

Crescent configurations in Normed Spaces
with D. King, S. J. Miller, E. A. Palsson, and C. Wahlenmayer
Integers, Vol. 20, #A96. arXiv

Local Properties via Color Energy Graphs and Forbidden Configurations
with Cosmin Pohoata and Adam Sheffer
SIAM Journal on Discrete Mathematics, Vol. 34, Jan 2020, pg. 177-187. arXiv

[Personal]