Ghazal Khalighinejad | Duke University

About Me

I am a PhD candidate in the Department of Computer Science at Duke University, fortunate to be advised by Prof. Bhuwan Dhingra, and previously advised by Sam Wiseman. Before Duke, I was an undergraduate student at Sharif University of Technology.

My research focuses on Natural Language Processing and Multimodal Large Language Models, with an emphasis on improving embedding methods for multimodal long-document retrieval and building LLM agents for scientific discovery. In the past, I’ve worked on structured information extraction from scientific articles, benchmarking the reasoning of multimodal models, and the algorithmic capabilities of transformers.

News

Summer 2026 — Document-as-Image Representations Fall Short for Scientific Retrieval accepted to COLM 2026.
Summer 2026 — Started my internship at Polymathic AI, working on AI agents for science.
Spring 2026 — New preprint: Document-as-Image Representations Fall Short for Scientific Retrieval.
Summer 2025 — Started my internship at Google.
Spring 2025 — Training Neural Networks as Recognizers of Formal Languages accepted at ICLR 2025.
Spring 2025 — MatViX accepted to NAACL 2025 (main conference, oral).
Fall 2024 — New preprint: MatViX, a benchmark for multimodal information extraction from long documents.
Summer 2024 — IsoBench accepted to the Conference on Language Modeling (COLM).
Summer 2024 — Started my internship at Adobe.
Spring 2024 — PNCExtract accepted to Findings of ACL 2024.
Spring 2024 — Introduced IsoBench, a benchmark for multimodal models with isomorphic inputs.

Older news

Fall 2023 — Approximating CKY with Transformers accepted to Findings of EMNLP 2023.
Summer 2023 — Received a full-tuition scholarship and research funding from the AI for Understanding and Designing Materials program at Duke.
Fall 2022 — Attended NeurIPS 2022 in New Orleans.
Spring 2022 — Undergraduate research paper accepted to ICALP 2022.
Fall 2021 — Started my PhD at Duke.

Selected Publications

COLM

Document-as-Image Representations Fall Short for Scientific Retrieval

Ghazal Khalighinejad, Raghuveer Thirukovalluru, Alexander H. Oh, Bhuwan Dhingra

Conference on Language Modeling (COLM), 2026.

PDF

NAACL

MatViX Multimodal Information Extraction from Visually Rich Articles

Ghazal Khalighinejad, Sharon Scott, Ollie Liu, Kelly L. Anderson, Rickard Stureborg, Aman Tyagi, Bhuwan Dhingra

NAACL 2025 (oral)

PDF Project Page

ICLR

Training Neural Networks as Recognizers of Formal Languages

Alexandra Butoi, Ghazal Khalighinejad, Anej Svete, Josef Valvoda, Ryan Cotterell, Brian DuSell

ICLR 2025

PDF Code

COLM

IsoBench Benchmarking Multimodal Foundation Models on Isomorphic Representations

Deqing Fu*, Ghazal Khalighinejad*, Ollie Liu*, and Bhuwan Dhingra, Dani Yogatama, Robin Jia, Willie Neiswanger

Conference on Language Modeling

PDF Project Page

ACL

Extracting Polymer Nanocomposite Samples from Full-Length Documents

Ghazal Khalighinejad, Defne Circi, LC Brinson, Bhuwan Dhingra

Findings of ACL, 2024.

PDF Code

EMNLP

Approximating CKY with Transformers

Ghazal Khalighinejad, Ollie Liu, Sam Wiseman

Findings of EMNLP, 2023.

PDF Code

ICALP

Galloping in fast-growth natural merge sorts

Elahe Ghasemi, Vincent Jugé, Ghazal Khalighinejad

International Colloquium on Automata, Languages, and Programming (ICALP), 2022.

PDF

Services

Conference Reviewer: ACL Rolling Review