Welcome to Jinxiang Meng’s personal homepage.

I am a Master’s student in Computer Science at the University of Chinese Academy of Sciences (UCASNJ), trained at the Institute of Automation, Chinese Academy of Sciences. I am advised by Professor Kang Liu.

My research focuses on building data and environment foundations for reliable general-purpose agents, especially data-intensive and long-horizon agents in real-world workflows. My interests include:

  • General agents and data agents
  • Long-horizon tasks, model capability optimization, and stable long-horizon training
  • Benchmark construction and execution-based evaluation
  • Synthetic data, environment synthesis, and post-training
  • Structured data reasoning, table reasoning, and Text-to-SQL

Open to opportunities: I am actively seeking 2027 Fall PhD opportunities or full-time job in LLM agents, data agents, long-horizon tasks, synthetic data, environment synthesis, and post-training. Please feel free to reach out if our interests align.

View CV Email

News

  • 2026.05: Our paper “DV-World: Benchmarking Data Visualization Agents in Real-World Scenarios” was accepted to ICML 2026.
  • 2026.04: Our paper “GATE: Graph-based Adaptive Tool Evolution Across Diverse Tasks” was accepted to ACL 2026.
  • 2026.02.14: Seed2.0 was released. My main contribution focused on improving the Seed model’s subjective data analysis capability through data construction and post-training data optimization for open-ended data analysis tasks.
  • 2026.01: Our paper “DAComp: Benchmarking Data Agents across the Full Data Intelligence Lifecycle” was accepted to ICLR 2026.
  • 2025.11: Our paper “TaREx: Reinforcement Learning for Code-Driven Table Reasoning” was accepted as an Oral at AAAI 2026.
  • 2025.06: We released the paper, code, and datasets for “Reasoning-Table: Exploring Reinforcement Learning for Table Reasoning”.
  • 2025.02: We released the paper and code for “GATE: Graph-based Adaptive Tool Evolution Across Diverse Tasks”.

Publications and Manuscripts

* denotes equal contribution.

DV-World: Benchmarking Data Visualization Agents in Real-World Scenarios

Jinxiang Meng, Shaoping Huang, Fangyu Lei, Jingyu Guo, Haoxiang Liu, Jiahao Su, Sihan Wang, Yao Wang, Enrui Wang, Ye Yang, Hongze Chai, Jinming Lv, Anbang Yu, Huangjing Zhang, Yitong Zhang, Yiming Huang, Zeyao Ma, Shizhu He, Jun Zhao, Kang Liu

ICML 2026

[PDF] [Code] [Project Page] [Dataset]

Not All Failures Are Equal: Pattern-Guided Fine-Tuning for Improving LLM Agents

Jinxiang Meng, Ruiyang Dong, Kuntao Yang, Jianwen Luo, Jiahui Liu, Fangyu Lei, Shizhu He, Jun Zhao, Kang Liu

Under Review

SheetForge: Scaling Spreadsheet Agents in Executable Environments

Jinxiang Meng, Fangyu Lei, Ruiyang Dong, Enrui Wang, Jianwen Luo, Shizhu He, Jun Zhao, Kang Liu

Under Review

DAComp: Benchmarking Data Agents across the Full Data Intelligence Lifecycle

Fangyu Lei*, Jinxiang Meng*, Yiming Huang, Junjie Zhao, Yitong Zhang, Jianwen Luo, Xin Zou, Ruiyi Yang, Wenbo Shi, Yan Gao, Shizhu He, Zuo Wang, Qian Liu, Yang Wang, Ke Wang, Jun Zhao, Kang Liu

ICLR 2026

[PDF] [Code] [Project Page] [Dataset]

TaREx: Reinforcement Learning for Code-Driven Table Reasoning

Fangyu Lei*, Jinxiang Meng*, Yiming Huang, Shizhu He, Jun Zhao, Kang Liu

AAAI 2026 Oral

[PDF]

DA-Omni: Scaling Training for Full-Stack Data Agents

Fangyu Lei*, Jinxiang Meng*, Jianwen Luo, Xingchang Yang, Kuntao Yang, Yitong Zhang, Shizhu He, Jun Zhao, Kang Liu

Under Review

Reasoning-Table: Exploring Reinforcement Learning for Table Reasoning

Fangyu Lei*, Jinxiang Meng*, Yiming Huang, Tinghong Chen, Yun Zhang, Shizhu He, Jun Zhao, Kang Liu

Under Review

[PDF] [Code] [Datasets]

GATE: Graph-based Adaptive Tool Evolution Across Diverse Tasks

Jianwen Luo*, Yiming Huang*, Jinxiang Meng, Fangyu Lei, Shizhu He, Xiao Liu, Shanshan Jiang, Bin Dong, Jun Zhao, Kang Liu

ACL 2026 Oral

[PDF] [Code]

VerifierArena: Benchmarking and Analyzing Universal Judges for Unverifiable Agentic Tasks

Fangyu Lei, Xingchang Yang, Kuntao Yang, Jinxiang Meng, Yitong Zhang, Shizhu He, Jun Zhao, Kang Liu

Under Review

Seed2.0 Model Card: Towards Intelligence Frontier for Real-World Complexity

ByteDance Seed

Technical Report, 2026

[PDF]

Education

  • 2024.09 - Present, Master’s student in Computer Science, University of Chinese Academy of Sciences (UCASNJ). Trained at the Institute of Automation, Chinese Academy of Sciences. Advisor: Prof. Kang Liu. GPA: 3.82/4.00; Rank: 1/35.
  • 2021.09 - 2022.07, Undergraduate visiting student, Information Management and Information Systems, Tianjin University, Tianjin.

Experience

Research Intern, Seed-Application Team, ByteDance

2025.08 - Present