Curriculum Vitae
Jingxi Qiu · Independent Researcher
Appointments
LLM Reliability, Evidence Verification & Selective RAG
Independent Research
- Built an initial SURE-RAG verifier with sampled answer–evidence checks and calibrated aggregation; this revealed artifacts in evidence-sufficiency evaluation and motivated a broader benchmark study.
- Developed NEI-CAP, a data-construction diagnostic framework showing that benchmark artifacts affect not only NEI detection but also SUPPORT/REFUTE confidence; ran controlled construction tests and multi-method ablations to audit and partially mitigate shortcut learning.
Volunteer Research Assistant (Offline RL)
Georgetown University
- Built a four-stage offline RL pipeline over 90M longitudinal patient-visit records for adaptive treatment and follow-up policy learning.
- Implemented KNN risk triage, Transformer risk estimation and Implicit Q-Learning over discrete treatment actions; held-out off-policy evaluation showed a 101.65% relative worst-case tail benefit (CVaR@5%) improvement over historical clinician-policy baselines.
Student Research Assistant (Machine Learning)
Georgetown University
- Led data analysis and modeling for multiple large-scale healthcare studies covering treatment access, telehealth adoption and educational equity.
- Analysed 125,000+ hematologic-malignancy patients and built a hierarchical multivariate logistic-regression model of disparities in CAR-T therapy access driven by the Social Vulnerability Index; contributed the analysis scripts and visualization pipeline to a study later published in The Oncologist.
Education
M.S., Data Science & Analytics
Georgetown University · Washington, D.C., USA
B.S., Finance, minor in Mathematics
Kean University · Union, NJ, USA
Publications
- 2026 · conference
Evidence Absence Is Not Evidence Insufficiency: Diagnosing NEI Construction Artifacts in Fact Verification
Jingxi Qiu, Zeyu Han, Cheng Huang
Findings of EMNLP
- 2026 · conference
SURE-RAG: Sufficiency and Uncertainty-Aware Evidence Verification for Selective Retrieval-Augmented Generation
Jingxi Qiu, Zeyu Han, Cheng Huang
IEEE PRAI
- 2026 · conference
CAT-Net: Image-to-Text for Medical Reports Using Adaptive Co-Attention and Triple-LSTM Module
Cheng Huang, Fan Gao, Yadi Liu, Zeyu Han, Jingxi Qiu, Jiahao Li, Wenbin Wei, Hao Tian, Yanteng Zhang, Qin Dai, Yongbin Yu†
IEEE EMBC
- 2025 · preprint
AttnBoost: Retail Supply Chain Sales Insights via Gradient Boosting Perspective
Yadi Liu, Xiaoli Ma, Muxin Ge, Zeyu Han, Jingxi Qiu, Ye Aung Moe, Yilan Shen, Wenbin Wei, Cheng Huang
arXiv
- 2025 · preprint
BioVessel-Net and RetinaMix: Unsupervised Retinal Vessel Segmentation from OCTA Images
Cheng Huang, Weizheng Xie, Fan Gao, Yutong Liu, Ruoling Wu, Zeyu Han, Jingxi Qiu, Xiangxiang Wang, Zhenglin Yang, Hao Wang, Yongbin Yu
arXiv (under review)
- 2022 · preprint
ProtSi: Prototypical Siamese Network with Data Augmentation for Few-Shot Subjective Answer Evaluation
Yining Lu, Jingxi Qiu, Gaurav Gupta
arXiv
Teaching
Teaching Assistant — DSAN 6700 ML Application Deployment
Georgetown University
- Supported instruction on deploying and productionizing machine-learning models, including Docker-based containerization and experiment management.
- Led hands-on support for Git, CI/CD pipelines, MLflow, locality-sensitive hashing and retrieval-augmented generation.
Skills
PyTorch · Transformers · LoRA / QLoRA · RLHF · Offline RL · Implicit Q-Learning · Off-policy Evaluation
BM25 · Reranking · LangChain · LangGraph · ReAct · Plan-and-Execute · Function Calling / Tool Use
Python · SQL · JavaScript · PySpark · Docker · Flask · Django · MLflow · GitHub Actions