Skip to main content

Selected research

Publications

A selection of papers and technical writing, from machine learning and experimentation to my earlier work in mathematics. For the full publication list and citation history, see my Google Scholar profile.

2024 · Technical report

Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone​

Marah Abdin et al. (including Amin Saied)

Introduces compact language models trained on curated and synthetic data, with Phi-3-mini demonstrating strong benchmark performance while remaining small enough to run locally on a phone.

2024 · Findings of NAACL

AGIEval: A Human-Centric Benchmark for Evaluating Foundation Models​

Wanjun Zhong, Ruixiang Cui, Yiduo Guo, Yaobo Liang, Shuai Lu, Yanlin Wang, Amin Saied, Weizhu Chen, Nan Duan

Findings of the Association for Computational Linguistics: NAACL 2024, pages 2299-2314.

Introduces a benchmark based on human examinations to evaluate foundation models' knowledge, reasoning, and problem-solving abilities, revealing where strong overall performance masks specific limitations.

2021 · ICLR

Economical Hyperparameter Optimization with Blended Search Strategy​

Chi Wang, Qingyun Wu, Silu Huang, Amin Saied

The Ninth International Conference on Learning Representations.

Combines global and local hyperparameter search, dynamically balancing exploration and computational cost to find high-quality model configurations with less time and resource consumption.

2020 · Technical article

Automated Metrics Calculation in a Dynamic Heterogeneous Environment​

Craig Boucher, Ulf Knoblich, Daniel Miller, Sasha Patotski, Amin Saied, Venky Venkateshaiah

Microsoft Experimentation Platform.

Presents a backend-independent metric language and code generation system that automate statistical analysis and data slicing, making experiment metrics reusable across products and compute platforms.

2018 · PhD thesis

Stable representation theory of categories and applications to families of (bi)modules over symmetric groups​

Amin Saied

Develops a framework for representation stability in families of symmetric-group bimodules, introducing PD-modules and connecting their theory to tensor algebras and the homology of partition lattices.

2017 · Mathematics

An efficient algorithm computing composition factors of T(V)⊗nT(V)^{\otimes n}​

Amin Saied

Turns the decomposition of tensor powers of free associative algebras into combinatorial counting problems, extending the range of computable representation-theoretic coefficients by more than 750-fold.

2016 · Mathematics

On the FI-module structure of Hi(Γn,s)H^i(\Gamma_{n,s})​

Amin Saied

Uses FI-modules to sharpen representation-stability bounds for the cohomology of groups arising from graph homotopy equivalences, and shows that their characters eventually follow polynomial formulas.