Peijie Qiu

I'm a PhD student at Washington University in St. Louis (WashU), advised by Prof. Aris Sotiras (now at University of Pennsylvania). My current and past research encompasses VLMs/LLMs/MLLMs, Reinforcement Learning from Human Feedback (RLHF), Generative Models (VAEs, GANs and Diffusion/Flow Models), (Multimodal) Representation Learning, Medical Image Analysis, and Sequential Modeling. I've also interned at Amazon as an applied scientist (hosted by Visual Search Team).

News

[02-23-2026] Two papers accepted to CVPR 2026🌟.
[05-23-2025] One paper accepted to MICCAI 2025🌟.
[05-01-2025] Two papers accepted to ICML 2025🌟.
[03-03-2025] I will be an intern at Amazon this summer, hosted by the Visual Search Team.
[12-09-2024] Two papers accepted to AAAI 2025🌟.

Research

	SafeUDPO: Safeguard Large Language Models With Utility-Calibrated Preference Optimization Author list is hidden for double-blind review. [preprint coming soon] TL;DR: A novel method to safeguard LLMs by calibrating the utility of listwise preference optimization.
	ImageRAGTurbo: Towards One-step Text-to-Image Generation with Retrieval-Augmented Diffusion Models Peijie Qiu, Hariharan Ramshankar, Arnau Ramisa, René Vidal, Vamsi Salaka, Rahul Bhagat, Amit Kumar K C [CVPR 2026] TL;DR: Accelerating text-to-image generation by retrieving and fusing relevant image features from a large database, enabling one-step generation with diffusion models.
	LLaDA-MedV: Exploring Large Language Diffusion Models for Biomedical Image Understanding Xuanzhao Dong, Wenhui Zhu, Xiwen Chen, Zhipeng Wang, Peijie Qiu, Shao Tang, Xin Li Yalin Wang (: co-first authors) [CVPR 2026; preprint coming soon] TL;DR:* The first dLLM for biomedical image understanding, improving both performance and efficiency compared to auto-regressive LLMs.
	DRA-GRPO: Your GRPO Needs to Know Diverse Reasoning Paths for Mathematical Reasoning Xiwen Chen, Wenhui Zhu, Peijie Qiu, Xuanzhao Dong, Hao Wang, Haiyu Wu, Huayu Li, Aristeidis Sotiras, Yalin Wang, Abolfazl Razi (: co-first authors) [arXiv; Submitted to ACL] TL;DR: A novel method to explore diverse reasoning paths for mathematical reasoning by downweighing the redundant paths, improving the performance of GRPO on MATH dataset by 7.5% (from 50.3% to 57.8%).
	Active Source-Free Cross-Domain and Cross-Modality Adaptation for Volumetric Medical Image Segmentation by Image Sensitivity and Organ Heterogeneity Sampling Jin Yang, Xiaobing Yu, Peijie Qiu, Daniel Marcus, Aristeidis Sotiras [MICCAI 2025] TL;DR: An active source-free 3D segmentation method that adapts across domains/modalities by sampling target volumes using sensitivity and heterogeneity.
	How Effective Can Dropout Be in Multiple Instance Learning ? Wenhui Zhu, Peijie Qiu, Xiwen Chen, Zhangsihao Yang, Aristeidis Sotiras, Abolfazl Razi, Yalin Wang (: co-first authors) [ICML 2025] TL;DR: A novel Dropout approach for MIL that significantly improves performance.
	Cracking Instance Jigsaw Puzzles: An Alternative to Multiple Instance Learning for Whole Slide Image Analysis Xiwen Chen, Peijie Qiu, Wenhui Zhu, Hao Wang, Huayu Li, Xuanzhao Dong, Xiaotong Sun, Xiaobing Yu, Yalin Wang, Abolfazl Razi, Aristeidis Sotiras (: co-first authors) [ICCV 2025] TL;DR: An alternative to permuation-invariant MIL for whole-slide image analysis by solving instance-level “jigsaw” puzzles.
	AgileFormer: Spatially Agile Transformer UNet for Medical Image Segmentation Peijie Qiu, Jin Yang, Sayantan Kumar, Soumyendu Sekhar Ghosh, Aristeidis Sotiras [Biomed. Signal Processing & Control] TL;DR: A roadmap to introduce spatially varying components into transformer-based UNets for segmentation.
	Multimodal Variational Autoencoder: a Barycentric View Peijie Qiu, Wenhui Zhu, Sayantan Kumar, Xiwen Chen, Xiaotong Sun, Jin Yang, Abolfazl Razi, Yalin Wang, Aristeidis Sotiras [AAAI 2025; Oral] TL;DR: A geometric view of multimodal VAEs that unifies existing probabilistic approaches and provides an information-theoretic view for probability mass allocation.
	Context-Aware Optimal Transport Learning for Retinal Fundus Image Enhancement Vamsi Krishna Vasa, Peijie Qiu, Wenhui Zhu, Yujian Xiong, Oana Dumitrascu, Yalin Wang (: co-first authors) [WACV 2025] TL;DR:* A novel context-aware optimal transport learning method for image-to-image translation.
	DGR-MIL: Exploring Diverse Global Representation in Multiple Instance Learning for Whole Slide Image Classification Wenhui Zhu, Xiwen Chen, Peijie Qiu, Aristeidis Sotiras, Abolfazl Razi, Yalin Wang (: co-first authors) [ECCV 2024] TL;DR: A novel method to explore diverse global representations for MIL-based whole slide image classification, improving the performance by 5.5% compared to the state-of-the-art methods.
	QCResUNet: Joint Subject-level and Voxel-level Segmentation Quality Prediction Peijie Qiu, Satrajit Chakrabarty, Phuc Nguyen, Soumyendu Sekhar Ghosh, Aristeidis Sotiras [Medical Image Analysis] [MICCAI 2023] TL;DR: A novel method for predicting the quality of medical image segmentation at both subject and voxel levels, improving the reliability of automated segmentation tools.
	TimeMIL: Advancing Multivariate Time Series Classification via a Time-aware Multiple Instance Learning Xiwen Chen, Peijie Qiu, Wenhui Zhu, Huayu Li, Hao Wang, Aristeidis Sotiras, Yalin Wang, Abolfazl Razi (: co-first authors) [ICML 2024] TL;DR: A novel time-aware MIL method for multivariate time series classification that models temporal dependencies and achieves state-of-the-art performance on several benchmarks.