Publications

Imaging-101: Benchmarking LLM Agents for Scientific Computational Imaging

Siyi Chen*, Jiahe Ying*, Yixuan Jia*, Yuxuan Gu*, Enze Ye, Weimin Bai, Zhijun Zeng, Shaochi Ren, Binhong Gao, Yubing Li, Tianhan Zhang, He Sun†

ICCP 2026 (Oral)

Paper Code

We introduce Imaging-101, a comprehensive benchmark evaluating LLM agents on scientific computational imaging tasks, revealing key failure modes including domain-specific scientific knowledge gaps, "scientific debugging atrophy," and lack of physics-grounded numerical intuition for units, scaling, and conditioning.

ForcingDAS: Unified and Robust Data Assimilation via Diffusion Forcing

Yixuan Jia†, Siyi Chen, Yida Pan, Xiao Li, Lianghe Shi, Chanyong Jung, Haijie Yuan, Ismail Alkhouri, Yue Cynthia Wu, Saiprasad Ravishankar, Jeffrey Fessler, Qing Qu

ICML 2026 Workshop (Spotlight)

Paper Code

We introduce ForcingDAS, a diffusion-forcing data assimilation framework that learns a joint-trajectory prior to reduce error accumulation, with a single trained model spanning the full filtering-to-smoothing spectrum at inference time.

Evaluating the Representation Space of Diffusion Models via Self-Supervised Principles

Xiao Li*, Yixuan Jia*, Zekai Zhang, Xiang Li, Lianghe Shi, Jinxin Zhou, Zhihui Zhu, Liyue Shen, Qing Qu†

ICML 2026 (Poster)

Paper Code

We propose the Invariant Contamination Ratio (ICR), a label-free metric based on Fisher-style invariance–residual decomposition to evaluate diffusion model feature quality across noise levels and training stages. ICR identifies optimal noise levels for feature extraction and serves as an early signal of memorization under limited data, without requiring labels or generation.

FlowDAS: A Stochastic Interpolant-Based Framework for Data Assimilation

Siyi Chen*, Yixuan Jia*, Qing Qu, He Sun†, Jeffrey Fessler

NeurIPS 2025 (Poster)

Paper Code

We introduce FlowDAS, a generative data assimilation framework that uses stochastic interpolants to learn observation-conditioned state transition dynamics from data, enabling step-by-step state estimation for stochastic dynamical systems without requiring known physical models.

Shorter SPECT Scans Using Self-Supervised Coordinate Learning to Synthesize Skipped Projection Views

Zongyu Li*, Yixuan Jia*†, Xiaojian Xu, Jason Hu, Yuni Dewaraja, Jeffrey Fessler

EJNMMI Physics | IMSI Computational Imaging 2024 Workshop (Oral)

Paper Code

We adapt the neural radiance field (NeRF) concept to SPECT imaging, enabling significant reduction in acquisition time (by 2×, 4×, or 8×) via self-supervised coordinate learning to synthesize skipped projection views.

Y90 SPECT Scatter Estimation and Voxel Dosimetry Using a Unified Deep Learning Framework

Yixuan Jia†, Zongyu Li, Azadeh Akhavanallaf, Jeffrey Fessler, Yuni Dewaraja

EJNMMI Physics | SNMMI 2023 (Oral)

Paper Code

We developed a unified three-stage deep learning framework for clinical Y90 SPECT imaging: CNN-based scatter estimation, SPECT reconstruction with scatter correction, and dose-rate map generation.

Data-Forcing Distillation: Restoring Diversity and Fidelity in Few-Step Video Generation

Siyi Chen, Shaowei Liu, Yixuan Jia, Zian Wang, Huan Ling, Qing Qu, Jun Gao†

ICML 2026 Workshop (Poster)

Paper Code

We propose Data-Forcing Distillation (DFD), a distillation method that restores both diversity and fidelity in few-step video generation, enabling efficient sampling without sacrificing generation quality.

MCLR: Improving Conditional Modeling in Visual Generative Models via Inter-Class Likelihood-Ratio Maximization

Xiang Li†, Yixuan Jia, Xiao Li, Jeffrey Fessler, Rongrong Wang, Qing Qu

ICML 2026 Workshop (Oral)

Paper Code

We propose MCLR, a principled alignment objective that maximizes inter-class likelihood-ratios during training, enabling diffusion models to achieve classifier-free guidance-like improvements under standard sampling without inference-time guidance. We further establish a formal equivalence between CFG and alignment-based objectives.

Deep Residual Learning Framework for Scatter Estimation in SPECT Imaging of Alpha Emitters: Application in ²¹¹At SPECT

Fardin Hussain, Yixuan Jia, Zongyu Li, Zhonglin Lu, Jeffrey Fessler, Yuni Dewaraja†

JNM Supplementary | SNMMI 2026 (Poster)

Paper Code

We develop a deep residual learning framework for scatter estimation in SPECT imaging of alpha emitters, demonstrated on ²¹¹At SPECT to enable more accurate quantitative imaging for targeted alpha therapy.

Segment Anything Model for SPECT (SAMS): Novel Implementation in SPECT Imaging for Tumor Segmentation

Zhonglin Lu, Zongyu Li, Yixuan Jia, Gefei Chen, Molly Roseland, Greta Mok, Yuni Dewaraja†

JNM Supplementary | SNMMI 2024 (Oral)

Paper Code

We adapt the Segment Anything Model (SAM) to SPECT imaging, enabling accurate tumor segmentation in nuclear medicine images.

Progress of Inertial Microfluidics in Principle and Application

Yixing Gou, Yixuan Jia, Peng Wang†, Changku Sun

Sensors

Paper Code

A comprehensive review of inertial microfluidics, covering the underlying physical principles and applications in particle manipulation, separation, and biomedical analysis.