Dat Nguyen

Short bio

I am a Postdoctoral Fellow at Harvard’s Programming Languages and Formal Methods groups and an incoming Postdoctoral Scientist at the Basis Research Institute.

My research focuses on program synthesis and probabilistic programming, with a track record in graph-based learning for code and documents. I completed my PhD at the University of Melbourne and previously worked at Cinnamon AI Lab on visually rich document information extraction.

At Harvard, I work on proof automation in Lean and causal systems for drug repurposing. At Basis, I contribute to MARA and R-ADA.

Research interests

Program synthesis and probabilistic programming
Graph-based learning for code and documents
Neuro-symbolic systems with LLMs and SMT
Reliable and explainable ML for software

Technical blogs

Grammars that generalize – Combining a small DSL with a neural network for domain-invariant bird recognition
Bayesian Synthesis – Bayesian synthesis of probabilistic programs for automatic data modeling
All posts

Project demos

	NeuroSymbolicDG PCFG over a spatial layout DSL as a domain-invariant classifier head for fine-grained bird recognition. code · blog · checkpoints
	VRDSynth (ISSTA '24) Synthesizing programs for multilingual visually rich document information extraction. code · paper
	Autumn.cpp An Autumn interpreter in C++ for MARA. code
	ExoPredicator (ICLR '26) Learning abstract models of dynamic worlds for robot planning. paper · openreview
	VirDA (TMLR '25) Reusing backbone for unsupervised domain adaptation with visual reprogramming. code · paper
	GNNInfer (ICSE '22, arXiv '24) Inferring properties of graph neural networks. paper
	FFL (ICSME '22) Fine-grained fault localization for student programs via syntactic and semantic reasoning. code · paper

Positions

Period	Role & Affiliation
2025 – present	Post-doctoral Fellow, Harvard SEAS & Basis Research Institute
2021 – 2024	PhD, School of Computing & Information Systems, University of Melbourne (Melbourne Research Scholarship)
2016 – 2021	AI Research Engineer, Cinnamon AI Lab

News

Jan 26, 2026	Co-authored paper “ExoPredicator” accepted at ICLR’26. Authors: Yichao Liang, Thanh Dat Nguyen, Cambridge Yang, Tianyang Li, Joshua B. Tenenbaum, Carl Edward Rasmussen, Adrian Weller, Zenna Tavares, Tom Silver, Kevin Ellis.
Jul 27, 2025	AutumnBench featured on the Basis Research Institute blog.
Mar 5, 2025	ArXiv: “A Systematic Survey on Debugging Techniques for Machine Learning Systems” (link).
Jan 15, 2025	Paper “VirDA” published in TMLR’25. Authors: Duc-Duy Nguyen, Dat Nguyen.
Dec 2, 2024	“Combining Induction and Transduction for Abstract Reasoning” won Best Paper at the ARC contest (arXiv).

Selected Publications [Full List]

ICLR

ExoPredicator: Learning Abstract Models of Dynamic Worlds for Robot Planning

Yichao Liang, Thanh-Dat Nguyen, Cambridge Yang, and 7 more authors

In 2026
arXiv

A Systematic Survey on Debugging Techniques for Machine Learning Systems

Thanh-Dat Nguyen, Haoye Tian, Bach Le, and 2 more authors

arXiv preprint arXiv:2503.03158 Mar 2025
TMLR

VirDA: Reusing Backbone for Unsupervised Domain Adaptation with Visual Reprogramming

Duc-Duy Nguyen, and Dat Nguyen

Transactions on Machine Learning Research Mar 2025
arXiv

Inferring Properties of Graph Neural Networks

Dat Nguyen, Hieu M. Vu, Cong-Thanh Le, and 4 more authors

arXiv preprint arXiv:2401.03790 Jan 2024
ISSTA

VRDSynth: Synthesizing Programs for Multilingual Visually Rich Document Information Extraction

Thanh-Dat Nguyen, Tung Do-Viet, Hung Nguyen-Duy, and 4 more authors

arXiv preprint arXiv:2407.06826 Jul 2024
arXiv

Combining Induction and Transduction for Abstract Reasoning

Wen-Ding Li, Keya Hu, Carter Larsen, and 11 more authors

arXiv preprint arXiv:2411.02272 Nov 2024
arXiv

Adversarial Attacks on Code Models with Discriminative Graph Patterns

Thanh-Dat Nguyen, Yang Zhou, Xuan Bach D. Le, and 2 more authors

arXiv preprint arXiv:2308.11161 Aug 2023
ICSME

FFL: Fine grained Fault Localization for Student Programs via Syntactic and Semantic Reasoning

Thanh-Dat Nguyen, Thanh Le-Cong, Duc-Minh Luong, and 4 more authors

In 2022 IEEE 38th International Conference on Software Maintenance and Evolution, Research Track Aug 2022
ICSE

Toward the Analysis of Graph Neural Networks

Thanh-Dat Nguyen, Thanh Le-Cong*, ThanhVu H. Nguyen, and 2 more authors

In 2022 IEEE/ACM 44th International Conference on Software Engineering: New Ideas and Emerging Results) Aug 2022

Abs

Graph Neural Networks (GNNs) have recently emerged as a robust framework for graph-structured data. They have been applied to many problems such as knowledge graph analysis, social networks recommendation, and even Covid19 detection and vaccine developments. However, unlike other deep neural networks such as Feed Forward Neural Networks (FFNNs), few analyses such as verification and property inferences exist, potentially due to dynamic behaviors of GNNs, which can take arbitrary graphs as input, whereas FFNNs which only take fixed size numerical vectors as inputs. This paper proposes an approach to analyze GNNs by converting them into FFNNs and reusing existing FFNNs analyses. We discuss various designs to ensure the scalability and accuracy of the conversions. We illustrate our method on a study case of node classification. We believe that our approach opens new research directions for understanding and analyzing GNNs.
ICPR

End-to-End Hierarchical Relation Extraction for Generic Form Understanding

Tuan-Anh Nguyen Dang, Duc Thanh Hoang, Quang Bach Tran, and 2 more authors

ICPR 2020 Aug 2020
MAPR

PCA-based 3D Facial Reenactment From Single Image

Thanh-Dat Nguyen, Tuan-Anh Nguyen Dang, and Viet Sang Dinh

In Aug 2020
BMVC

End-to-End Information Extraction by Character-Level Embedding and Multi-Stage Attentional UNet

Tuan Anh Nguyen Dang, and Thanh-Dat Nguyen

British Machine Vision Conference (BMVC) Aug 2019
Non-local DenseNet for plant CLEF 2019 contest

Thanh-Dat Nguyen, George Quénot, and Lorraine Goeuriot

CEUR-Workshop Aug 2019