Mengzhao Jia

Robotics and Agentic AI

Portrait of Mengzhao Jia

I am a fourth-year Ph.D. student in Computer Science and Engineering at the University of Notre Dame, advised by Prof. Meng Jiang. Before starting my Ph.D., I received my master’s and bachelor’s degrees from Shandong University. I previously interned at Orby AI in Mountain View and Tencent AI Lab in Seattle.

My research focuses on improving the reasoning capabilities of multimodal foundation models through reasoning-oriented data curation and rubric-based reinforcement learning. Recently, I have expanded my research to agentic AI for robotics, exploring how general-purpose agents can control robots through visual reasoning, programming, and physical feedback.

I am open to internships and full-time roles in robotics and agentic AI. Please get in touch.

Publications

01Preprint 2026

Agent as Policy for Robotic Manipulation

Mengzhao Jia*, Yang Lin*, Xixin Zhang*, Zhihan Zhang, Xiaobai Liu, Meng Jiang

* Core contributors

Agent as Policy (AGP) lets a general purpose agent control a physical robot through visual reasoning, runtime programming, and feedback from execution. It performs assembly, block construction, dice flipping, targeted throwing, and bimanual towel folding without task or environment specific training.

LabApprentice robot preparing a hydrogel at the laboratory bench

02Project 2026

LabApprentice: Self-Improving Wet-Lab Robotics Harness

LabApprentice explores how a robotic coding agent can accumulate and reuse practical tools through physical experiments. It writes and retains programs for pouring, heating, dispensing, and instrument control, then brings these tools together to prepare and measure hydrogels.

MMTutorBench overview

03ACL 2026

MMTutorBench: The First Multimodal Benchmark for AI Math Tutoring

Tengchao Yang*, Sichen Guo*, Mengzhao Jia, Jiaming Su, Yuanyang Liu, Zhihan Zhang, Meng Jiang

* Equal Contribution

A benchmark for evaluating multimodal models on mathematical tutoring across insight discovery, operation formulation, and operation execution.

AutoRubric method overview

04ACL 2026 (Findings)

AutoRubric: Rubric-Based Generative Rewards for Faithful Multimodal Reasoning

Mengzhao Jia, Zhihan Zhang, Ignacio Cases, Zheyuan Liu, Meng Jiang, Peng Qi

A rubric-driven reward framework for improving the accuracy and faithfulness of multimodal reasoning.

Groupwise Ranking Reward overview

05EMNLP 2026 (Findings)

Prioritizing the Best: Incentivizing Reliable Multimodal Reasoning by Rewarding Beyond Answer Correctness

Mengzhao Jia, Zhihan Zhang, Meng Jiang

A groupwise ranking reward that favors reliable, verifier-passed multimodal reasoning trajectories beyond final-answer correctness.

Leopard model overview

06TMLR 2025

Leopard: A Vision Language Model for Text-Rich Multi-Image Tasks

Mengzhao Jia, Wenhao Yu, Kaixin Ma, Tianqing Fang, Zhihan Zhang, Siru Ouyang, Hongming Zhang, Dong Yu, Meng Jiang

A vision-language model and instruction data for reasoning over text-rich, multi-image inputs.

MLLMU-Bench overview

07NAACL 2025

Protecting Privacy in Multimodal Large Language Models with MLLMU-Bench

Zheyuan Liu, Guangyao Dou, Mengzhao Jia, Zhaoxuan Tan, Qingkai Zeng, Yongle Yuan, Meng Jiang

A benchmark for evaluating multimodal machine unlearning and privacy protection in large language models.

MultiChartQA benchmark overview

08NAACL 2025

MultiChartQA: Benchmarking Vision-Language Models on Multi-Chart Problems

Zifeng Zhu*, Mengzhao Jia*, Zhihan Zhang, Lang Li, Meng Jiang

* Equal Contribution

A benchmark for multi-hop, comparative, and sequential reasoning across multiple charts.

Query-oriented micro-video summarization model overview

09IEEE TPAMI 2024

Query-Oriented Micro-Video Summarization

Mengzhao Jia, Yinwei Wei, Xuemeng Song, Teng Sun, Min Zhang, Liqiang Nie

A multimodal framework for generating concise, query-oriented summaries of micro-videos to support retrieval.

Describe-then-Reason training and inference pipeline

10Preprint 2024

Describe-then-Reason: Improving Multimodal Mathematical Reasoning through Visual Comprehension Training

Mengzhao Jia, Zhihan Zhang, Wenhao Yu, Fangkai Jiao, Meng Jiang

A two-step training approach that improves multimodal mathematical reasoning through visual comprehension training.

Counterfactual data augmentation for multimodal sarcasm detection

11AAAI 2024

Debiasing Multimodal Sarcasm Detection with Contrastive Learning

Mengzhao Jia, Can Xie, Liqiang Jing

A contrastive framework that reduces spurious textual bias for robust out-of-distribution multimodal sarcasm detection.

Work Experience

  1. Mar — Sep 2025

    Research Intern

    Orby AI · Mountain View, CA

  2. May — Sep 2024

    Research Intern

    Tencent AI Lab · Seattle, WA

Education

  1. 2023 — Present

    Ph.D. in Computer Science and Engineering

    University of Notre Dame, advised by Prof. Meng Jiang.

  2. 2020 — 2023

    M.S. in Computer Science and Engineering

    Shandong University

  3. 2016 — 2020

    B.Eng. in Electronic Science and Technology

    Shandong University