1. Introduction
We propose Pulsar-VLM, a visual reasoning model for pulsar candidate identification, built on pretrained multimodal large language models (MLLMs). The input to Pulsar-VLM includes diagnostic subplots (i.e., frequency-phase plots and DM curves) along with task instructions. By integrating the visual features of the diagnostic subplots with semantic textual descriptions, Pulsar-VLM is able to perform geometric morphological reasoning on the diagnostic plots in a manner analogous to how astronomers interpret them.
Pulsar-VLM maintains high performance across diagnostic plots from commonly used radio telescopes, despite morphological differences, and thus provides substantial benefits to the pulsar research community.
2. Evaluation Dataset
Below are the pulsar candidate datasets used in this work:
CRAFTS Dataset: Hugging Face Repository
Provided by Dr. Pei Wang, Institute for Frontiers in Astronomy and Astrophysics, Beijing Normal University. Email: wangpei@nao.cas.cnGPPS Dataset: Official Website
GC FANS Dataset: Official Website
Diagnostic plots generated with PRESTO, from which we identified 91 pulsars. Our processed dataset is available at: Hugging Face RepositoryFAST Dataset: GitHub Repository
GBT-350 Dataset: Official Website
GBNCC Dataset: Official Website
SGAN Dataset: GitHub Repository
CHIRSS Dataset: Official Website
LOTAAS Dataset: Official Website
3. Evaluation Results
| Dataset | FN | Recall |
|---|---|---|
| CRAFTS Dataset | 3 | 98.8% |
| GC FANS Dataset | 1 | 98.9% |
| GPPS Dataset | 3 | 99.8% |
| GBT-350 Dataset | 0 | 100% |
| GBNCC Dataset | 10 | 94.2% |
| LOTAAS Dataset | 0 | 100.0% |
| CHIRSS Dataset | 0 | 100.0% |
| FAST Dataset | 1 | 99.7% |
| SGAN Dataset | 577 | 94.5% |