AstroSpecLM: A Spectrum-Language Model for Evidence-Grounded Astronomical Spectral Analysis
Abstract
Astronomical spectra encode rich physical information, but drawing scientific conclusions from spectral features typically requires expert interpretation. This paper presents AstroSpecLM, a spectrum-language model that connects one-dimensional DESI spectra with Qwen3-4B to answer questions and provide explanations grounded in spectral evidence. Instead of generating question-answer pairs directly from templates or raw catalog fields, we first distill each spectrum into a compact set of catalog- and spectrum-derived facts, then use these facts as references to generate instruction-following conversations. The resulting model is competitive with specialist supervised baselines on classification and redshift estimation, while additionally producing natural-language explanations that reference specific spectral features. Our results indicate that grounding a language model in one-dimensional scientific spectra is feasible, and that fact-mediated instruction data yields a model capable of both prediction and explanation.
Related
- Automated, auditable dating of the Babylonian Astronomical Diaries with language models and astronomical tools
- The Great Chat in the Sky: Aligning Exotic Astronomical Data to Language with AstroBridge
- Evidence of a Low-Energy Cutoff in the Injected Spectrum of a Pulsar Wind Nebula
- Executable verification through formalized expert reasoning in astronomical spectroscopy
- The Astronomical Hub: A Unified Ecosystem for Modern Astronomical Research
BibTeX
@misc{shi2026astrospeclm,
title = {{AstroSpecLM: A Spectrum-Language Model for Evidence-Grounded Astronomical Spectral Analysis}},
author = {Jinghang Shi and Yanxia Zhang and Ali Luo and Changhua Li and Xiao Kong},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.07102},
eprint = {2609.07102},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.07102},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - AstroSpecLM: A Spectrum-Language Model for Evidence-Grounded Astronomical Spectral Analysis
AU - Jinghang Shi
AU - Yanxia Zhang
AU - Ali Luo
AU - Changhua Li
AU - Xiao Kong
PY - 2026
DA - 2026/09/07
PB - arXiv
DO - 10.48550/arxiv.2609.07102
UR - https://arxiv.org/abs/2609.07102
AB - Astronomical spectra encode rich physical information, but drawing scientific conclusions from spectral features typically requires expert interpretation. This paper presents AstroSpecLM, a spectrum-language model that connects one-dimensional DESI spectra with Qwen3-4B to answer questions and provide explanations grounded in spectral evidence. Instead of generating question-answer pairs directly from templates or raw catalog fields, we first distill each spectrum into a compact set of catalog- and spectrum-derived facts, then use these facts as references to generate instruction-following conversations. The resulting model is competitive with specialist supervised baselines on classification and redshift estimation, while additionally producing natural-language explanations that reference specific spectral features. Our results indicate that grounding a language model in one-dimensional scientific spectra is feasible, and that fact-mediated instruction data yields a model capable of both prediction and explanation.
ER -