![]() |
葛凌岳 Ling-Yue Ge |
Sept 2025 - present: Admitted to study for a M.Sc. degree in Nanjing University without entrance examination under the supervision of Professor Lan-Zhe Guo.
Sept 2021 - June 2025: Received my B.Sc. degree of Artificial Intelligence from School of Artifitial Intelligence, at Nanjing University.
Self-Evolving Agents: Developing autonomous agents that iteratively refine their roles, memory, tool-use strategies, and decision policies through interaction feedback, self-reflection, and environment-grounded evaluation.
MLLM Visual Planning: Investigating how multimodal large language models perceive structured visual environments, compose spatial-temporal plans, and make constraint-aware decisions for complex visual planning tasks.
Roles with Rails: Contract-Preserving Role Evolution in Multi-Agent Structured Reasoning.
Ling-Yue Ge, Lan-Zhe Guo.
arXiv 2026
[Paper]
[PDF]
We propose SERO, a contract-preserving role evolution framework for multi-agent structured reasoning. Frozen role pools cannot adapt to new tasks, while unconstrained role generation breaks the implicit contracts among agents (capability, communication, validation, aggregation, and output). SERO introduces an Edit Gate that only commits role edits which preserve these contracts, allowing the multi-agent system to adapt to novel tasks while maintaining reliable coordination.
MapTab: Are MLLMs Ready for Multi-Criteria Route Planning in Heterogeneous Graphs?
Ziqiao Shang*, Lingyue Ge*, Zi-Jian Cheng, Shi-Yu Tian, Zhenyu Huang, Wenbo Fu, Weiming Wu, Yang Chen, Xiangwen Zhang, Yulan Hu, Bin Liu, Yu-Feng Li, Lan-Zhe Guo.
arXiv 2026
[Paper]
[PDF]
[Code]
(* denotes equal contribution)
We introduce MapTab, a benchmark and systematic study examining whether multimodal large language models (MLLMs) are ready for multi-criteria route planning over heterogeneous graphs. MapTab evaluates models on planning routes that satisfy multiple, often competing criteria across visually-grounded heterogeneous map structures, revealing the strengths and limitations of current MLLMs in constraint-aware spatial planning.
First Prize of People's Scholarship, NJU in Nanjing, 2023 and 2024
Second Prize of People's Scholarship, NJU in Nanjing, 2022
Teaching Assistant, Introduction to Artificial Intelligence (Nanjing University, For Undergraduate Students), Fall, 2025