Michael Hu

I am a third-year PhD student at the NYU Center for Data Science, working with Kyunghyun Cho and Tal Linzen. I work on algorithms that optimize the training data of language models. I am supported by an NSF Graduate Research Fellowship.

In my spare time, I enjoy cooking, running, and playing basketball.

Previously, I completed a BSE at Princeton CS, where I spent two lovely years working with Karthik Narasimhan and Tom Griffiths.

news

Mar 26, 2025	Gave talks on “Between Circuits and Chomsky” at École Normale Supérieure CoML and FLaNN.
Mar 15, 2025	“Aioli: A Unified Optimization Framework for Language Model Data Mixing” and “How to visualize training dynamics in neural networks” accepted to ICLR 2025.
Feb 27, 2025	New preprint: “Between Circuits and Chomsky: Pre-pretraining on Formal Languages Imparts Linguistic Biases”.
Jul 17, 2024	New preprint: “The importance of human-scale language modeling for psycholinguistics.”
Nov 21, 2023	“Latent State Models of Training Dynamics” accepted to TMLR.

selected publications

2025

Preprint

Between Circuits and Chomsky: Pre-pretraining on Formal Languages Imparts Linguistic Biases

Michael Y. Hu, Jackson Petty, Chuan Shi, and 2 more authors

Preprint, 2025

arXiv
ICLR

Aioli: A Unified Optimization Framework for Language Model Data Mixing

Mayee F. Chen*, Michael Y. Hu*, Nicholas Lourie, and 2 more authors

ICLR, 2025

arXiv Website

2024

Preprint

Bigger is not always better: The importance of human-scale language modeling for psycholinguistics

Ethan Wilcox, Michael Y. Hu, Aaron Mueller, and 6 more authors

Preprint, 2024

Website

2023

TMLR

Latent State Models of Training Dynamics

Michael Y. Hu, Angelica Chen, Naomi Saphra, and 1 more author

Transactions on Machine Learning Research, 2023

Website

2021

NeurIPS

Safe Reinforcement Learning with Natural Language Constraints

Tsung-Yen Yang, Michael Hu, Yinlam Chow, and 2 more authors

NeurIPS, 2021

Abs arXiv Website

While safe reinforcement learning (RL) holds great promise for many practical applications like robotics or autonomous cars, current approaches require specifying constraints in mathematical form. Such specifications demand domain expertise, limiting the adoption of safe RL. In this paper, we propose learning to interpret natural language constraints for safe RL. To this end, we first introduce HazardWorld, a new multi-task benchmark that requires an agent to optimize reward while not violating constraints specified in free-form text. We then develop an agent with a modular architecture that can interpret and adhere to such textual constraints while learning new tasks. Our model consists of (1) a constraint interpreter that encodes textual constraints into spatial and temporal representations of forbidden states, and (2) a policy network that uses these representations to produce a policy achieving minimal constraint violations during training. Across different domains in HazardWorld, we show that our method achieves higher rewards (up to11x) and fewer constraint violations (by 1.8x) compared to existing approaches. However, in terms of absolute performance, HazardWorld still poses significant challenges for agents to learn efficiently, motivating the need for future work.