Pinqiao Wang

Pinqiao Wang

PhD Candidate in Data Science
University of Virginia, advised by Dr. Sheng Li

I am a PhD Candidate at the RISE Lab, School of Data Science, UVA. My research focuses on developing ML/AI methods to understand agent behaviors for designing efficient algorithmic-based systems of LLM and to create reliable agent-based applications. Specific interests include Agentic AI, LLM Reasoning, Multimodal LLMs, and AI Applications.

Email Scholar GitHub CV

📣 News

💼 Experience

PhD Research Assistant
RISE Lab, School of Data Science, University of Virginia
Charlottesville, VA · Aug 2025 – Present
Advisor: Prof. Sheng Li. Research on Agentic AI, LLM reasoning, and Multi-Modality LLM.
Research Lead (part-time)
AI4Finance Foundation
Remote · Jan 2024 – Present
Leading Finance LLM development and evaluation. FinRobot Pro ranked Top 50 on GitHub trending.
Published first-author paper at ACM ICAIF 2024 (oral, top 5%). Developed multi-layer debate RL algorithm for equity research report generation.
AI Research Intern
Agam Capital Management, LLC
Teaneck, NJ · Jun 2024 – Aug 2024
Designed RAG + MoE inference pipeline achieving 85% accuracy in automated code translation and Q&A generation.
Research Intern
Baidu, Inc. · NLP Lab
Shenzhen, China · May 2023 – Aug 2023
Enhanced ERNIE model for legal document summarization, achieving 15% improvement in accuracy over internal benchmarks.

📝 Publications

Preprints
  • Under Review
    Multi-Agent Debate: A Unified Agentic Framework for Tabular Anomaly Detection
    Pinqiao Wang, Hanzhang Yuan, Sheng Li
    Paper Code
  • Under Review
    VITAL: Visual–Text Invariance Tests for Text-Prior Reliance in Multimodal Language Models
    Pinqiao Wang, Daiqing Qi, Tianlong Wang, Wenhao Zhang, Lehan Yang, Sheng Li
    Paper Code
  • Under Review
    LANTERN: Learning A Pathology Vision Foundation Model via Text-Supervised Disentangled Representation
    Wenhao Zhang, Zhongliang Zhou, John Kang, Edward T. Richardson, John B. Wojcik, Carol E. Peña, Qiaohui Zhou, Lehan Yang, Tianlong Wang, Pinqiao Wang, Wei Liu, Sheng Li
    Paper Code
Peer-Reviewed
  • EMNLP 2026
    You Shouldn’t Have Asked: A Pragmatics-Inspired Taxonomy for Evaluating LLM Refusals
    Ruoxuan Li*, Pinqiao Wang*, Sheng Li, Cameron Robert Jones
    Paper Code Equal Contribution
  • TMLR 2026
    ItinBench: Benchmarking Planning Across Multiple Cognitive Dimensions with Large Language Models
    Tianlong Wang, Pinqiao Wang, Weili Shi, Sheng Li
    Paper Code
  • ICML 2026 Workshop
    Evaluating Spatial World Modeling in Video Generators via 3D Camera Trajectory Generation
    Tianlong Wang, Lehan Yang, Yuheng Liu, Hanzhang Yuan, Wenhao Zhang, Pinqiao Wang, Yifan Li, Yu Kong, Sheng Li
    Paper Code
  • ACM ICAIF 2024
    FinRobot: AI Agent for Equity Research and Valuation with Large Language Models
    Pinqiao Wang, Tianyu Zhou*, Yilin Wu*, Hongyang Yang
    Paper Code Google Scholar Oral (Top 5%)
* Equal contribution.

🔬 Projects

Multi-Agent Systems · 2025–present

MAD

Multi-Agent Debate — a unified agentic framework for tabular anomaly detection. Treats detector disagreement as a first-class signal, resolved through a mathematically grounded coordination layer with LLM-based critics.

Paper Code
Multimodal LLM · 2025–present

CODA

Conflict-Aware Dynamic Arbitration — investigating how multimodal LLMs handle text-visual conflicts during inference. Built on Bayesian precision weighting and semantic entropy.

Paper Code
AI Agent · ACM ICAIF 2024

FinRobot

AI agent platform for equity research and valuation with LLMs. Multi-layer debate RL algorithm with chain-of-thought prompting for producing full equity research reports. FinRobot Pro ranked Top 50 on GitHub trending.

Paper Code
Multi-Agent Systems · UVA

AI Teaching Assistant

Agentic system for teaching and administrative work at UVA's School of Data Science. Working with Prof. Prince Afriyie and Prof. Gianluca Guadagni.

Demo Code
LLM Benchmark · 2025

ItinBench

Benchmarking LLM planning capabilities across multiple cognitive dimensions — evaluating spatial, temporal, and constraint-based reasoning in travel itinerary generation.

Paper
More to come

New projects in progress...

🏆 Awards

📋 Service

Reviewer:  ICML 2026,  ACL 2025

🎓 Education

Ph.D. in Data Science
University of Virginia
Charlottesville, VA · Expected May 2030
M.A. in Statistics with Honor
Columbia University
New York, NY · December 2024
B.S. in Statistics & Analytics, and Economics Highest Distinction
University of North Carolina at Chapel Hill
Chapel Hill, NC · May 2023

🎵 Personal

Outside the lab, I compose and produce music. The creative process shares more with research than one might expect — iterating on ideas, embracing happy accidents, and knowing when something finally clicks.

More about my interests and life outside of research will appear here. Feel free to reach out if you'd like to chat about research, music, or anything in between.

Music

Original compositions

01
Echoes On The Highway
MP3
02
Not a kid
MP3
03
where to now?
MP3