Publications

Phenotypic prediction of missense variants via deep contrastive learning

Jun Wen, Sihang Zeng, Clara-Lea Bonzel, Shilpa Nadimpalli Kobren, Jiangchuan Du, Yi Chai, Hao Wang, Meng Zhu, Siwei Chen, Fangwei Leng, Harrison G. Zhang, Katherine P. Liao, Kelly Cho, Isaac S. Kohane, Marinka Zitnik, Alexandre C. Pereira, Jun S. Liu, Tianxi Cai

Nature Biomedical Engineering

PheMART connects missense variants to 4,179 clinical phenotypes in a shared metric space, combining protein language models, interaction networks, medical knowledge graphs, and EHR-derived signal to support rare disease diagnosis.

Genomic classification to predict survival in metastatic prostate cancer: development of Somatic Tumor Risk Assessment for Overall Survival–Prostate (STRATOS-P)

Martin W. Schoen, Jiannong Li, Sihang Zeng, Heena Desai, Ryan Hausler, Candace L. Haroldsen, Lukas Owens, Luca F. Valle, Ruth B. Etzioni, Timothy R. Rebbeck, Brent S. Rose, Michael J. Kelley, R. Bruce Montgomery, Nicholas G. Nickols, Matthew B. Rettig, Kosj Yamoah, Kara N. Maxwell, Isla P. Garraway

JCO Precision Oncology

A prognostic classifier built from comprehensive genomic profiling of 7,201 veterans with metastatic prostate cancer in the VHA National Precision Oncology Program.

The role of Whole Health in enhancing tobacco cessation outcomes for veterans: a retrospective cohort study

Sihang Zeng, Scott S. Coggeshall, Ethan W. Rosser, Stephanie L. Taylor, Diana J. Burgess, Gang Luo, Steven B. Zeliadt

Journal of General Internal Medicine

Links participation in Whole Health services to measurable tobacco cessation outcomes across the Veterans Health Administration.

CoRTEx: contrastive learning for representing terms via explanations with applications on constructing biomedical knowledge graphs

Huaiyuan Ying, Zhengyun Zhao, Yang Zhao, Sihang Zeng, Sheng Yu

Journal of the American Medical Informatics Association 31(9), 1912–1920

Explanation-augmented contrastive learning that sharpens biomedical term representations for large-scale knowledge graph construction.

A feasibility study of 2-D microwave thorax imaging based on the supervised descent method

Haolin Zhang, Maokun Li, Fan Yang, Shenheng Xu, Yan Yin, Hongyu Zhou, Yubo Yang, Sihang Zeng, Jianchong Shao

Electronics 10(3), 352

Learning-based 2-D microwave thorax imaging via the supervised descent method for fast structural reconstruction.

CONTEXTOR: contextualized high-order contrastive learning

Ze Cai, Hanzhe Liang, Sihang Zeng, Binbin Zhou, Jun Wen

International Conference on Machine Learning (ICML)

Recasts high-order relation inference as a dynamic query–response process, contextualizing candidate entities through asymmetric conditional modulation.

MARTI: a framework for multi-agent LLM systems reinforced training and inference

Kaiyan Zhang, Kai Tian, Runze Liu, Sihang Zeng, Xuekai Zhu, Guoli Jia, Yuchen Fan, Xingtai Lv, Yuxin Zuo, Che Jiang, et al.

International Conference on Learning Representations (ICLR)

An open framework unifying reinforced training and inference for multi-agent LLM systems.

ReviewRL: towards automated scientific review with reinforcement learning

Sihang Zeng, Kai Tian, Kaiyan Zhang, Junqi Gao, Runze Liu, Sa Yang, Jingxuan Li, Xinwei Long, Jiaheng Ma, Biqing Qi, Bowen Zhou

Conference on Empirical Methods in Natural Language Processing (EMNLP) Main conference

Trains review generation with verifiable rewards so that automated reviews stay grounded in the paper rather than drifting into fluent generality.

TrajSurv: learning continuous latent trajectories from electronic health records for trustworthy survival prediction

Sihang Zeng, Lucas Jing Liu, Jun Wen, Meliha Yetisgen, Ruth Etzioni, Gang Luo

Machine Learning for Healthcare Conference (MLHC) PMLR 298

Learns a continuous-time latent trajectory per patient, so survival predictions come with an inspectable account of how risk accumulated.

Traj-CoA: patient trajectory modeling via chain-of-agents for lung cancer risk prediction

Sihang Zeng, Yujuan Fu, Sitong Zhou, Zixuan Yu, Lucas Jing Liu, Jun Wen, Matthew Thompson, Ruth Etzioni, Meliha Yetisgen

NeurIPS GenAI4Health Workshop

A chain of LLM agents reads a patient's notes in order and passes forward a compressed timeline, making zero-shot lung cancer risk prediction traceable to source evidence.

UW-BioNLP at ChemoTimelines 2025: thinking, fine-tuning, and dictionary-enhanced LLM systems for chemotherapy timeline extraction

Tianmai M. Zhang, Zhaoyi Sun, Sihang Zeng, Chenxi Li, Neil F. Abernethy, Barbara D. Lam, Fei Xia, Meliha Yetisgen

Clinical NLP Workshop (ACL) pp. 40–56 1st place, subtask 2

Winning system for generating patient chemotherapy timelines from raw clinical notes, combining fine-tuning, chain-of-thought, and dictionary-enhanced retrieval.

Scalability of LLM-based multi-agent systems for scientific code generation: a preliminary study

Yuru Wang, Kaiyan Zhang, Kai Tian, Sihang Zeng, Xingtai Lv, Ning Ding, Biqing Qi, Bowen Zhou

MathNLP Workshop (EMNLP)

Shows a minimalist actor–critic pair can beat a reasoning model at equal compute while cutting token cost by 75%.

UltraMedical: building specialized generalists in biomedicine

Kaiyan Zhang, Sihang Zeng, Ermo Hua, Ning Ding, Zhang-Ren Chen, Zhiyuan Ma, Haoxin Li, Ganqu Cui, Biqing Qi, Xuekai Zhu, et al.

NeurIPS Datasets and Benchmarks Track Spotlight

A large biomedical instruction dataset and model suite that reaches domain specialization without giving up general capability.

Large language models as biomedical hypothesis generators: a comprehensive evaluation

Biqing Qi, Kaiyan Zhang, Kai Tian, Haoxiang Li, Zhang-Ren Chen, Sihang Zeng, Ermo Hua, Jinfang Hu, Bowen Zhou

Conference on Language Modeling (COLM)

Evaluates biomedical hypothesis generation on date-partitioned background–hypothesis pairs to control for contamination.

Large language models are zero shot hypothesis proposers

Biqing Qi, Kaiyan Zhang, Haoxiang Li, Kai Tian, Sihang Zeng, Zhang-Ren Chen, Bowen Zhou

NeurIPS Instruction Tuning Workshop

Tests whether language models can propose scientifically meaningful biomedical hypotheses without task-specific training.

Automatic biomedical term clustering by learning fine-grained term representations

Sihang Zeng, Zheng Yuan, Sheng Yu

BioNLP Workshop (ACL) pp. 91–96

Fine-grained term representations that separate near-synonyms well enough to cluster biomedical terminology automatically.

Traj-Evolve: a self-evolving multi-agent system for patient trajectory modeling in lung cancer early detection

Sihang Zeng, Matthew Thompson, Ruth Etzioni, Meliha Yetisgen

arXiv:2606.02812

Pairs an experience pool of retrieved similar patients with RL-optimized agent coordination, beating nine baselines over five years of patient history.

NatureBench: can coding agents match the published SOTA of Nature-family papers?

Yuru Wang, Lejun Cheng, Yuxin Zuo, Sihang Zeng, Bingxiang He, Che Jiang, Junlin Yang, Yuchong Wang, Kaikai Zhao, et al.

arXiv:2606.24530

Ninety tasks drawn from peer-reviewed Nature papers. The best agent exceeds published results on 17.8% of them, mostly by reducing science to familiar supervised learning.

EpiEvolve: self-evolving agents for streaming pandemic forecasting under regime shifts

Yiming Lu, Sihang Zeng, Zhengxu Tang, Max Lau, Fei Liu, Wei Jin

arXiv:2606.05513

Keeps a fixed LLM forecaster but adapts through episodic memory, halving the time to recover after a variant regime shift.

TrajOnco: a multi-agent framework for temporal reasoning over longitudinal EHR for multi-cancer early detection

Sihang Zeng, Young Won Kim, Wilson Lau, Ehsan Alipour, Ruth Etzioni, Meliha Yetisgen, Anand Oka

arXiv:2604.10386

A training-free multi-agent LLM framework with memory that matches or beats supervised baselines across 15 cancer types (AUC 0.64–0.80) while producing reviewable evidence.

Self-improving agents in the era of experience: a survey of self- to meta-evolution

Che Jiang, Jincheng Zhong, Yu Fu, Kai Tian, Junlin Yang, Kaikai Zhao, Yuchong Wang, Tianwei Luo, Weizhi Wang, Yuxin Zuo, Guoli Jia, Xingtai Lv, Dianqiao Lei, Sihang Zeng, et al.

Preprint

Surveys agents that improve themselves, and the meta-layer that decides which parts of the improvement process should change.

A survey of reinforcement learning for large reasoning models

Kaiyan Zhang, Yuxin Zuo, Bingxiang He, Youbang Sun, Runze Liu, Che Jiang, Yuchen Fan, Kai Tian, Guoli Jia, Pengfei Li, Yu Fu, Xingtai Lv, Yuchen Zhang, Sihang Zeng, et al.

arXiv:2509.08827

A broad survey of RL for reasoning models: foundational components, open problems, training resources, and where scaling runs into walls.

CaSBRE: causality-inspired semi-supervised biomedical relation extraction

Sihang Zeng, Jun Wen, Jiangchuan Du, Jing Qian, Tianxi Cai, Hao Wang

Under review

Disentangles causal from spurious features so relation extraction stays robust with scarce labels and unseen entities.

Hierarchical pretraining for biomedical term embeddings

Bryan Cai, Sihang Zeng, Yucong Lin, Zheng Yuan, Doudou Zhou, Lu Tian

arXiv:2307.00266

Hierarchy-aware pretraining that encodes graded semantic relatedness between biomedical terms.

BIOS: an algorithmically generated biomedical knowledge graph

Sheng Yu, Zheng Yuan, Jun Xia, Shengxuan Luo, Huaiyuan Ying, Sihang Zeng, Jingyi Ren, Hongyi Yuan, Zhengyun Zhao, Yucong Lin, et al.

arXiv:2203.09975

A biomedical knowledge graph built by machine learning rather than manual curation, at a scale manual curation cannot reach.

Predicting multi-cancer risk from EHR data using multi-agent LLMs

Sihang Zeng, Young Won Kim, Wilson Lau, Ehsan Alipour, Ruth B. Etzioni, Meliha Yetisgen, Anand Oka, Jayashree Nanduri

Journal of Clinical Oncology ASCO Annual Meeting

STRATOS-P clinical model for prognosis and selection of patients with metastatic hormone-sensitive prostate cancer for intermittent therapy

Martin W. Schoen, Joshua Gruber, Jason M. Doherty, David B. Eaton Jr., Sihang Zeng, Lukas Owens, et al.

Journal of Clinical Oncology ASCO Annual Meeting

Development of an oncology generative AI foundation model trained on more than a million longitudinal patient journeys across the United States

Wilson Lau, Ehsan Alipour, Young Won Kim, Sihang Zeng, Anand Oka, Jayashree Nanduri

Journal of Clinical Oncology ASCO Annual Meeting

MSR67 — Zero-shot lung cancer risk prediction from longitudinal electronic health records with a chain-of-agents framework

Sihang Zeng, Young Won Kim, Wilson Lau, Ehsan Alipour, Ruth Etzioni, Meliha Yetisgen, Anand Oka, Jayashree Nanduri

Value in Health ISPOR

MSR68 — Characterizing oncology patient journeys and health state transitions using a data-driven Markov transition matrix in large-scale electronic health records

Young Won Kim, Wilson Lau, Ehsan Alipour, Sihang Zeng, Anand Oka, Jayashree Nanduri

Value in Health ISPOR

MSR172 — Can a generative patient journey foundation model alleviate the burden of cancer screening?

Wilson Lau, Ehsan Alipour, Young Won Kim, Sihang Zeng, Anand Oka, Jayashree Nanduri

Value in Health ISPOR

RWD140 — Patient journey foundational model for scalable imputation of missing units of measurement in electronic health records data

Ehsan Alipour, Wilson Lau, Young Won Kim, Sihang Zeng, Anand Oka, Jayashree Nanduri

Value in Health ISPOR

Adapting the Global Burden of Disease Healthcare Access and Quality Index for the Veterans Health Administration: a feasibility study

Sihang Zeng, Christine Wilson, Gang Luo, Steven Zeliadt

AcademyHealth Annual Research Meeting

Population-level tobacco cessation outcomes associated with implementing Whole Health at the Veterans Health Administration

Sihang Zeng, Scott Coggeshall, Ethan Rosser, Stephanie Taylor, Diana Burgess, Gang Luo, Steven Zeliadt

AcademyHealth Annual Research Meeting

Towards trustworthy modeling of patient trajectory with longitudinal electronic health records

Sihang Zeng

PhD dissertation, University of Washington

Advised by Ruth Etzioni and Meliha Yetisgen. Degree conferred June 2026.