publications

2025

1. GALA: A Multi-LLM-Agent Framework for Graph Classification

   2025

   Submitted to ACL 2025. Preprint available upon request.

   Abs Bib

   We propose GALA, a novel multi-agent framework for graph classification tasks. The framework supports end-to-end, config-driven auto-training using PyTorch and Python multiprocessing for parallel experiments, achieving a 5% accuracy improvement and 3x speedup over baseline pipelines.

   @article{xiang2025gala,
     title = {GALA: A Multi-LLM-Agent Framework for Graph Classification},
     year = {2025},
     note = {Submitted to ACL 2025. Preprint available upon request.},
   }

2024

1. RegiViz — A Tool for Generating and Visualizing Cancer Regimens

   In Brown University CSCI 2370 Proceedings, 2024

   Abs Bib PDF Website

   Cancer regimens guide patients through chemotherapy, but existing visualizations lack automation, personalization, and clarity. RegiViz, a web-based tool, automates cancer regimen visualization using a fine-tuned large language model (LLM) trained on the HemOnc database. It features intuitive visual encodings for drug schedules, types, and treatment timelines, with holiday-aware adjustments and patient-doctor feedback integration. User studies show RegiViz improves efficiency and clarity over existing regimen visualizations.

   @inproceedings{xiang2024regiviz,
     title = {RegiViz — A Tool for Generating and Visualizing Cancer Regimens},
     year = {2024},
     booktitle = {Brown University CSCI 2370 Proceedings},
   }