I am Songwei Dong (董松玮), a third-year Ph.D. student at the University of Virginia, where I am fortunate to be advised by Professor Cong Shen. Previously, I received my B.S. degree in Artificial Intelligence in 2024 from the University of Science and Technology of China.
My research is on augmenting large language models with external knowledge, tools, and memory, spanning LLM agents (multi-agent systems and agent memory) and retrieval-augmented generation (RAG). I also work on vision-language models (VLMs), with broader interests in multimodal learning, coreset selection, and LLM efficiency.
Warning
Problem: The current name of your GitHub Pages repository ("Solution: Please consider renaming the repository to "
http://".
However, if the current repository name is intended, you can ignore this message by removing "{% include widgets/debug_repo_name.html %}" in index.html.
Action required
Problem: The current root path of this site is "baseurl ("_config.yml.
Solution: Please set the
baseurl in _config.yml to "Education
-
University of VirginiaDepartment of Computer EngineeringPh.D. StudentAug. 2024 - present -
University of Science and Technology of ChinaB.S. in Artificial IntelligenceSep. 2020 - Jun. 2024
Experience
-
CableLabsResearch InternJun. 2026 - Sep. 2026
Honors & Awards
-
Excellent Student Scholarship, USTC2021 - 2023
Selected Publications (view all )
SpecMind: Enabling Spectrum Intelligence via Multi-Agent Hybrid Retrieval-Augmented Generation
A multi-agent hybrid RAG system that routes spectrum questions to specialised agents over license records, regulations, and proceeding filings, then composes one grounded answer.
SpecMind: Enabling Spectrum Intelligence via Multi-Agent Hybrid Retrieval-Augmented Generation
A multi-agent hybrid RAG system that routes spectrum questions to specialised agents over license records, regulations, and proceeding filings, then composes one grounded answer.
The Past Is Prologue: A Plug-in Controller for Selective Updates in Sequentially Evolving LLM Memory
A plug-in controller that decides whether each candidate memory update is worth deploying, adding 2.7 to 4.6 accuracy points on top of existing memory updaters.
The Past Is Prologue: A Plug-in Controller for Selective Updates in Sequentially Evolving LLM Memory
A plug-in controller that decides whether each candidate memory update is worth deploying, adding 2.7 to 4.6 accuracy points on top of existing memory updaters.
Is One Score Enough? Rethinking the Evaluation of Sequentially Evolving LLM Memory
A diagnostic framework for LLM memory that separates online utility, generalization, backward transfer, and forgetting, showing that headline accuracy often hides substantial forgetting.
Is One Score Enough? Rethinking the Evaluation of Sequentially Evolving LLM Memory
A diagnostic framework for LLM memory that separates online utility, generalization, backward transfer, and forgetting, showing that headline accuracy often hides substantial forgetting.