2026
MedQPA-Gen: Medical Question Proposing and Answering for Report Generation
Weijie Liang*, Xiyue Zhu*#, Ruike Zhu, C. Li, C. Tang, Z. Liu, Z. Gong, S. Luo, Y. Li, Volodymyr Kindratenko (* equal contribution, # corresponding author)
Findings of the Association for Computational Linguistics (ACL Findings) 2026 Medical imaging VLM / LLM Agents RL post-training Reward design
MedQPA is a question-driven agentic evaluator for medical reports; MedQPA-Gen optimizes report generators against it with reflective prompting, iterative DPO, and GRPO on Qwen2-VL-7B, reaching an 80% human-preference win rate over the base model.
MedQPA-Gen: Medical Question Proposing and Answering for Report Generation
Weijie Liang*, Xiyue Zhu*#, Ruike Zhu, C. Li, C. Tang, Z. Liu, Z. Gong, S. Luo, Y. Li, Volodymyr Kindratenko (* equal contribution, # corresponding author)
Findings of the Association for Computational Linguistics (ACL Findings) 2026 Medical imaging VLM / LLM Agents RL post-training Reward design
MedQPA is a question-driven agentic evaluator for medical reports; MedQPA-Gen optimizes report generators against it with reflective prompting, iterative DPO, and GRPO on Qwen2-VL-7B, reaching an 80% human-preference win rate over the base model.
2025
Turbocharging Web Automation: The Impact of Compressed History States
Xiyue Zhu, Peng Tang, Haofu Liao, Srikar Appalaraju
Findings of the Association for Computational Linguistics (ACL Findings) 2025 VLM / LLM Agents
A learned history compressor distills each verbose past web state into a fixed-length representation, cutting context length by ~70% and improving step accuracy by 2–5% on Mind2Web and WebLINX.
Turbocharging Web Automation: The Impact of Compressed History States
Xiyue Zhu, Peng Tang, Haofu Liao, Srikar Appalaraju
Findings of the Association for Computational Linguistics (ACL Findings) 2025 VLM / LLM Agents
A learned history compressor distills each verbose past web state into a fixed-length representation, cutting context length by ~70% and improving step accuracy by 2–5% on Mind2Web and WebLINX.
Introducing 3D Representation for Volume-to-Volume Translation via Score Fusion
Xiyue Zhu, Dou Hoon Kwark, Ruike Zhu, Kaiwen Hong, Yiqi Tao, Shirui Luo, Yudu Li, Zhi-Pei Liang, Volodymyr Kindratenko
International Conference on Machine Learning (ICML) 2025 Diffusion & generative models Medical imaging Video generation
Learn to fuse perpendicularly trained 2D diffusion models in score space with a lightweight 3D network, bringing a true 3D representation to medical volume-to-volume translation; 3–10% better on 3D MRI super-resolution and video inverse problems.
Introducing 3D Representation for Volume-to-Volume Translation via Score Fusion
Xiyue Zhu, Dou Hoon Kwark, Ruike Zhu, Kaiwen Hong, Yiqi Tao, Shirui Luo, Yudu Li, Zhi-Pei Liang, Volodymyr Kindratenko
International Conference on Machine Learning (ICML) 2025 Diffusion & generative models Medical imaging Video generation
Learn to fuse perpendicularly trained 2D diffusion models in score space with a lightweight 3D network, bringing a true 3D representation to medical volume-to-volume translation; 3–10% better on 3D MRI super-resolution and video inverse problems.
2024
Accurate Feature Extraction from Historical Geologic Maps Using Open-Set Segmentation and Detection
A. Saxton, J. Dong, A. Bode, N. Jaroenchai, R. Kooper, Xiyue Zhu, et al.
Geosciences 2024 Segmentation & detection
Open-set segmentation and detection for extracting features from historical geologic maps.
Accurate Feature Extraction from Historical Geologic Maps Using Open-Set Segmentation and Detection
A. Saxton, J. Dong, A. Bode, N. Jaroenchai, R. Kooper, Xiyue Zhu, et al.
Geosciences 2024 Segmentation & detection
Open-set segmentation and detection for extracting features from historical geologic maps.
2023
MapPrior: Bird's-Eye View Map Layout Estimation with Generative Models
Xiyue Zhu, Vlas Zyrianov, Zhijian Liu, Shenlong Wang
IEEE/CVF International Conference on Computer Vision (ICCV) 2023 Autonomous driving Diffusion & generative models
A generative prior over semantic map layouts on top of a discriminative BEV perception model: better accuracy, realism (MMD), and uncertainty awareness (ECE) on nuScenes.
MapPrior: Bird's-Eye View Map Layout Estimation with Generative Models
Xiyue Zhu, Vlas Zyrianov, Zhijian Liu, Shenlong Wang
IEEE/CVF International Conference on Computer Vision (ICCV) 2023 Autonomous driving Diffusion & generative models
A generative prior over semantic map layouts on top of a discriminative BEV perception model: better accuracy, realism (MMD), and uncertainty awareness (ECE) on nuScenes.
2022
Learning to Generate Realistic LiDAR Point Clouds
Vlas Zyrianov, Xiyue Zhu, Shenlong Wang
European Conference on Computer Vision (ECCV) 2022 Autonomous driving Diffusion & generative models
LiDARGen generates realistic LiDAR point clouds by score-based denoising in the equirectangular view and supports conditional generation without retraining.
Learning to Generate Realistic LiDAR Point Clouds
Vlas Zyrianov, Xiyue Zhu, Shenlong Wang
European Conference on Computer Vision (ECCV) 2022 Autonomous driving Diffusion & generative models
LiDARGen generates realistic LiDAR point clouds by score-based denoising in the equirectangular view and supports conditional generation without retraining.