Reza Shirkavand

PhD Candidate in Computer Science, University of Maryland

I am a PhD candidate in the Department of Computer Science at the University of Maryland, College Park, advised by Dr. Heng Huang.

I work on making large models adaptive: models that understand their own limitations, select the best internal or external expert, use their failures to select or generate what to learn next, and adapt efficiently and safely when data is scarce.

News

Jun 2026 Flash-BoN is accepted at ECCV 2026.
Jan 2026 IDIOMoE is accepted at ICLR 2026.
Dec 2025 ToMoE is accepted at TMLR, to be presented at ICML 2026.
Sep 2025 CSCR is accepted at NeurIPS 2025 as a Spotlight.
Sep 2025 Bilevel-ZOFO is accepted at NeurIPS 2025.
Jun 2025 ARGUS is accepted at ICCV 2025.
Feb 2025 FT-Unlearn is accepted at CVPR 2025.
Jan 2025 APTP is accepted at ICLR 2025.

Research

2026

Flash-BoN: Instant Drafts for Inference-Time Scaling in Diffusion Models

Ruchit Rawal, Reza Shirkavand, Sayak Paul, Yuxin Wen, Heng Huang, Yizheng Chen, Tom Goldstein, and Gowthami Somepalli

ECCV 2026

Paper Code Project Abstract

Catalog-Native LLM: Speaking Item-ID dialect with Less Entanglement for Recommendation

Reza Shirkavand, Xiaokai Wei, Chen Wang, Zheng Hui, Heng Huang, Michelle Gong

ICLR 2026

Paper Code Abstract

ToMoE: Converting Dense Large Language Models to Mixture-of-Experts through Dynamic Structural Pruning

Shangqian Gao, Ting Hua, Reza Shirkavand, Chi-Heng Lin, Zhen Tang, Zhengao Li, Longge Yuan, Fangyi Li, Zeyu Zhang, Alireza Ganjdanesh, Lou Qian, Xu Jie, Yen-Chang Hsu

TMLR @ICML 2026

Paper Code Abstract

2025

Cost-Aware Contrastive Routing for LLMs

Reza Shirkavand, Shangqian Gao, Peiran Yu, Heng Huang

NeurIPS 2025 ▸ Spotlight

Paper Code Abstract

Bilevel ZOFO: Bridging Parameter-Efficient and Zeroth-Order Techniques for Efficient LLM Fine-Tuning and Meta-Training

Reza Shirkavand, Peiran Yu, Qi He, Heng Huang

NeurIPS 2025

Paper Code Abstract

ARGUS: Hallucination and Omission Evaluation in Video-LLMs

Ruchit Rawal, Reza Shirkavand, Heng Huang, Gowthami Somepalli, Tom Goldstein

ICCV 2025

Paper Code Data Abstract

2024

Efficient Fine-Tuning and Concept Suppression for Pruned Diffusion Models

Reza Shirkavand, Peiran Yu, Shangqian Gao, Gowthami Somepalli, Tom Goldstein, Heng Huang

CVPR 2025

Paper Code Abstract

Not All Prompts Are Made Equal: Prompt-based Pruning of Text-to-Image Diffusion Models

Reza Shirkavand*, Alireza Ganjdanesh*, Shangqian Gao, Heng Huang

ICLR 2025 * equal contribution

Paper Code Abstract

From Pixels to Prose: A Large Dataset of Dense Image Captions

Vasu Singla, Kaiyu Yue, Sukriti Paul, Reza Shirkavand, Mayuka Jayawardhana, Alireza Ganjdanesh, Heng Huang, Abhinav Bhatele, Gowthami Somepalli, Tom Goldstein

ArXiv

Paper Data Abstract

2023

Deep Prompt Tuning for Graph Transformers

Reza Shirkavand, Heng Huang

ArXiv

Paper Code Abstract

Incomplete Multimodal Learning for Complex Brain Disorders Prediction

Reza Shirkavand, Liang Zhan, Heng Huang, Li Shen, Paul M Thompson

ArXiv

Paper Code Abstract

2021

Dementia Severity Classification under Small Sample Size and Weak Supervision in Thick Slice MRI

Reza Shirkavand, Sana Ayromlou, Soroush Farghadani, Maedeh-sadat Tahaei, Fattane Pourakpour, Bahareh Siahlou, Zeynab Khodakarami, Mohammad H Rohban, Mansoor Fatehi, Hamid R Rabiee

ArXiv

Paper Code Abstract

Teaching

Advanced Machine Learning Topics

Co-Instructor · University of Maryland

Fall 2026 · Spring 2026 · Spring 2025 · Spring 2024

Lectures on transformers and diffusion models.