Lihong Wang

Multimodal AI · Research & Engineering

Lihong Wang

Multimodal Algorithm Engineer at Ant Phecda Lab, Ant Group

Ant Phecda Lab and 蚂蚁天玑实验室 logo
Ant Phecda Lab蚂蚁天玑实验室Ant Digital Technologies · Ant Group

My work focuses on multimodal reasoning, reinforcement learning, and agentic vision-language models, with an emphasis on reliable training data and interpretable reasoning processes.

CVPR 2026 ViRCICLR 2025 Diff3DSAnt Group Multimodal AI
Portrait of Lihong Wang by the lake at sunset
Based in Hangzhou · Open to research collaboration

Recent updates

News

Research releases, publications, and milestones.

Dataset release

We released AerialVLN-Verified and AerialVLN-S-Verified, two high-quality datasets curated with ViRC-7B and 150 hours of human verification.

AerialVLN-VerifiedAerialVLN-S-Verified
CVPR 2026

ViRC was accepted to CVPR 2026. It is an agentic multimodal model for visual interleaved mathematical reasoning.

PaperCode
ICLR 2025

Diff3DS was accepted to ICLR 2025, introducing differentiable curve rendering for view-consistent 3D sketch generation.

ProjectPaper

Selected work

Publications

Research spanning multimodal mathematical reasoning, financial reasoning data, and differentiable rendering.

ViRC progressive training strategy and agentic visual reasoning framework
CVPR2026

ViRC: Enhancing Visual Interleaved Mathematical CoT with Reason Chunking

Lihong Wang, Liangqi Li, Weiwei Feng, Jiamin Wu, Changtao Miao, Tieru Wu, Rui Ma, Bo Zhang, Zhe Li

PaperCodeModelDataset
Agentar-DeepFinance-100K systematic data synthesis and verification pipeline
arXiv preprint2025

Agentar-DeepFinance-100K: A Large-Scale Financial Dataset via Systematic Chain-of-Thought Synthesis Optimization

Xiaoke Zhao, Zhaowen Zhou, Lin Chen, Lihong Wang, Zhiyi Huang, Kaiyuan Zheng, Yanjun Zheng, Xiyang Du, Longfei Liao, Jiawei Liu, Xiang Qi, Bo Zhang, Peng Zhang, Wei Wang, Zhe Li

PaperCodeDataset
Diff3DS differentiable rendering pipeline for view-consistent 3D sketches
ICLR2025

Diff3DS: Generating View-Consistent 3D Sketch via Differentiable Curve Rendering

Yibo Zhang, Lihong Wang, Changqing Zou, Tieru Wu, Rui Ma

PaperProjectICLR
View all publications (6)

Path

Experience Highlights

View full experience →
2026.06 — Present

Multimodal Algorithm Engineer

Ant Phecda Lab (蚂蚁天玑实验室), Ant Digital Technologies, Ant Group

Building multimodal reasoning systems and reinforcement learning pipelines for vision-language models.

2023 — 2026

M.E. in Artificial Intelligence

Jilin University

Multimodal reasoning research in the Intelligent Content Learning Group.

Recognition

Selected Awards

View all awards →
2023.09

National Second Prize

The Interpretable Auxiliary Diagnosis of Scoliosis Based on Learnable B-Splines

5th China Graduate AI Innovation Competition

2022.07

First Prize, South China Division

Privacy Protection System Based on Image Encryption and Invisible Watermarking

2022 C4-Network Technology Challenge, A-ST Track

2022.08

National Third Prize

Multimedia Data Security Protection System for Medical Diagnosis

15th National College Student Information Security Contest

Connect

Let’s build more capable multimodal systems.

For research discussions, open-source collaboration, or professional opportunities, feel free to get in touch.

Start a conversation