Jikai Jin

Jikai Jin

Ph.D. student

Stanford university

Biography

Welcome to my personal website! I am a third year Ph.D. student in The Institute for Computational and Mathematical Engineering (ICME), Stanford university, working with Prof. Vasilis Syrgkanis. Prior to joining Stanford, I obtained my bachelor degree in computational mathematics in the School of Mathematical Sciences, Peking University, fortunately having Prof. Liwei Wang as my research advisor. I am broadly interested in any research questions that combine theoretical insights with real-world impact. My recent research interests include (1) understanding fundamental limits of ML-empowered estimation methods and (2) applying statistical tools to enhance LLM evaluation. If you share similar interests, feel free to contact me via email or Wechat. Download my resumé.

Interests
  • ML and Statistics: ML-empowered statistical estimation
  • ML and Causality: identifiability theory and algorithms
  • Statistics and LLM Eval: enhance statistical rigor and provide novel insights
Education

  • Ph.D. in Computational and Mathematical Engineering, 2023 - 2027 (expected)

    Stanford University

  • BSc in Computational Mathematics, 2019 - 2023

    Peking University

  • High School Diploma, 2017 - 2019

    No.2 High School of East China Normal University

Recent News

All news»

Feb. 2026 New paper on prescriptive scaling law posted on ArXiv!

Feb. 2026 New paper Adaptive Exploration for Latent-State Bandits posted on ArXiv! This is a joint work with my Meta intern mentor Kenneth Hung and his colleagues.

Dec. 2025 New paper Sharp Structure-Agnostic Lower Bounds for General Linear Functional Estimation with Vasilis Syrgkanis posted on ArXiv!

Dec. 2025 Attending NeurIPS 2025 at San Diego! Super happy to see lots of old friends there.

Jul. 2025 New paper It’s Hard to Be Normal: The Impact of Noise on Structure-agnostic Estimation with Lester Mackey and Vasilis Syrgkanis posted on ArXiv!

Recent Publications

Quickly discover relevant content by filtering publications.
Hanlin Zhang, Jikai Jin, Vasilis Syrgkanis, Sham Kakade (2026). Prescriptive Scaling Reveals the Evolution of Language Model Capabilities. arXiv.

PDF DOI ArXiv Blog Dataset

Jikai Jin, Kenneth Hung, Sanath Kumar Krishnamurthy, Baoyi Shi, Congshan Zhang (2026). Adaptive Exploration for Latent-State Bandits. arXiv.

Cite DOI ArXiv

Ayush Sawarni, Jikai Jin, Justin Whitehouse, Vasilis Syrgkanis (2026). Policy Learning with Abstention. In AISTATS 2026.

Cite ArXiv

Jikai Jin, Vasilis Syrgkanis (2026). Sharp Structure-Agnostic Lower Bounds for General Linear Functional Estimation. arXiv.

Cite DOI ArXiv

Jikai Jin, Lester Mackey, Vasilis Syrgkanis (2025). It's Hard to Be Normal: The Impact of Noise on Structure-agnostic Estimation . In NeurIPS 2025.

ArXiv

Jikai Jin, Vasilis Syrgkanis, Sham Kakade, Hanlin Zhang (2025). Discovering Hierarchical Latent Capabilities of Language Models via Causal Representation Learning. In arXiv preprint arXiv:2506.10378.

PDF Cite ArXiv

Jiayi Sheng, Luna Lyu, Jikai Jin, Tony Xia, Alex Gu, James Zou, Pan Lu (2025). Solving Inequality Proofs with Large Language Models. In NeurIPS 2025 (Datasets and Benchmarks track, spotlight).

PDF Cite ArXiv Website Twitter

Jikai Jin, Vasilis Syrgkanis (2025). Structure-agnostic Optimality of Doubly Robust Learning for Treatment Effect Estimation. In COLT 2025.

Cite ArXiv Poster Slides Blog

Jikai Jin, Vasilis Syrgkanis (2024). Learning Causal Representations from General Environments: Identifiability and Intrinsic Ambiguity. In NeurIPS 2024 (spotlight).

Cite ArXiv Slides

Kaifeng Lyu, Jikai Jin, Zhiyuan Li, Simon S. Du, Jason D. Lee, Wei Hu (2023). Dichotomy of Early and Late Phase Implicit Biases Can Provably Induce Grokking. In ICLR 2024.

Cite ArXiv

Jikai Jin, Zhiyuan Li, Kaifeng Lyu, Simon S. Du, Jason D. Lee (2023). Understanding Incremental Learning of Gradient Descent -- A Fine-grained Analysis of Matrix Sensing. In ICML 2023.

PDF Cite ArXiv Poster Slides

Jikai Jin, Yiping Lu, Jose Blanchet, Lexing Ying (2022). Minimax Optimal Kernel Operator Learning via Multilevel Training. In ICLR 2023 (spotlight).

PDF Cite ArXiv Slides Poster

Binghui Li, Jikai Jin, Han Zhong, John E. Hopcroft, Liwei Wang (2022). Why Robust Generalization in Deep Learning is Difficult: Perspective of Expressive Power. In NeurIPS 2022.

PDF Cite ArXiv

Jikai Jin, Suvrit Sra (2022). Understanding Riemannian Acceleration via a Proximal Extragradient Framework. In COLT 2022.

PDF Cite ArXiv

Jikai Jin, Bohang Zhang, Haiyang Wang, Liwei Wang (2021). Non-convex Distributionally Robust Optimization: Non-asymptotic Analysis. In NeurIPS 2021.

PDF Cite ArXiv

Bohang Zhang, Jikai Jin, Cong Fang, Liwei Wang (2020). Improved analysis of clipping algorithms for non-convex optimization. In NeurIPS 2020.

PDF Cite ArXiv

Blog

Prescriptive Scaling for Language Models
Given a pre-training compute budget, we can predict the best downstream performance strong post-training can reliably attain, and detect when that boundary shifts.
Jikai Jin
Last updated on Feb 9, 2026 13 min read Blog

Experience

 
 
 
 
 
advised by Prof. Liwei Wang (Peking University)
Undergraduate Research Intern
advised by Prof. Liwei Wang (Peking University)
Feb 2020 – Present Beijing, China
Work on machine learning theory.

Awards and Honors

All awards and honors»

Jun. 2023 Huaixin Scholar, BICMR

Jun. 2023 Peking University Excellent Graduate.

Jan. 2023 Sensetime Scholarship (awarded to 30 Chinese undergraduate students in the field of AI).

Oct. 2022 Qin Wanshun-Jin Yunhui Scholarship, Peking University.

Oct. 2022 Merit Student, Peking University.

Oct. 2021 Exceptional award for academic innovation, Peking University.

Jun. 2021 Elite undergraduate training program of Applied Mathematics and Statistics.

May 2021 Bronze Medal, S.T. Yau College Student Mathematics Competition, probability and statistics individual.

Dec. 2020 Qin Wanshun-Jin Yunhui Scholarship, Peking University.

Oct. 2020 Yizheng Scholarship, Peking University.

Feb. 2019 Silver Medal, 11th Romania Masters of Mathematics.

Oct. 2018 First Prize (ranked No.6), Chinese Mathematical Olympiad (CMO).

Oct. 2017 First Prize (ranked No.13), Chinese Mathematical Olympiad (CMO).

Contact