Skip to content
DraconiticResearch
Search Sources API
194,164 studies31 sourcesUpdated 4 min ago
Peer-reviewedCC BYThe Astrophysical Journal Supplement Series

Mephisto: Self-improving Large Language Model–based Agents for Automated Interpretation of Multiband Galaxy Observations

  • Zechang Sun1
  • Yuan-Sen Ting2
  • Yaobo Liang3
  • Nan Duan3
  • Song Huang1
  • Zheng Cai1
  1. 1Tsinghua University
  2. 2The Ohio State University
  3. 3Microsoft Research Asia
Open at sourceDOICite

Abstract

Astronomical research has long relied on human expertise to interpret complex data and formulate scientific hypotheses. In this study, we introduce Mephisto—a multiagent collaboration framework powered by large language models (LLMs) that emulates human-like reasoning to analyze multiband galaxy observations. Mephisto interfaces with the CIGALE codebase (a library of spectral energy distribution models) to iteratively refine physical models against observational data. It conducts deliberate reasoning via a tree search, accumulates knowledge through self-play, dynamically updates its knowledge base, and is validated across diverse galaxy populations—including the James Webb Space Telescope’s recently discovered “little red dot” galaxies. We show that Mephisto demonstrates the ability to infer the physical properties of galaxies from multiband photometry, positioning it as a research copilot for astronomers. Unlike prior black-box machine learning approaches in astronomy, Mephisto offers a transparent, human-aligned reasoning process that integrates with existing research practices. This work underscores the possibility of LLM-driven agent-based research for astronomy, establishes a foundation for fully automated, end-to-end artificial intelligence (AI)-powered scientific workflows, and opens up new avenues for AI-augmented discoveries in astronomy.

Related

  1. Searching for Black Hole Candidates in Quiescence by Using Multiband Observations in Globular Cluster M22 (NGC 6656)
    Yu-Jing 雨婧 Xu 徐, Wei-Min 为民 Gu 顾, Yuanqi 媛琪 Liu 刘 +3
    The Astrophysical JournalPeer-reviewedCC BY 4.0Full textSolar & stellar physicsCrossref
  2. Multiband Color Monitoring of 3I/ATLAS through Ground-Based Relay Observations
    Ariel Graykowski, Bryce Bolin, Laura-May Abron +140
    arXivPreprintCC BY 4.0Full textGalaxiesPlanetary science & exoplanetsarXiv
  3. A Grid of Contribution Functions to Help the Interpretation of Time-resolved Observations of Brown Dwarfs
    Myrla Phillippe, Theodora Karalidi, Elena Manjavacas +4
    The Astrophysical JournalPeer-reviewedCC BY 4.0Full textSolar & stellar physicsCrossref
  4. A Grid of Contribution Functions to Help the Interpretation of Time-resolved Observations of Brown Dwarfs
    Myrla Phillippe, Theodora Karalidi, Elena Manjavacas +4
    arXivPreprintCC BY 4.0Full textPlanetary science & exoplanetsSolar & stellar physicsarXiv
  5. Automated optimization of solar energetic particle acceleration and transport parameters against five-spacecraft observations of the 9 October 2021 event
    Hiroyuki Tako Ishikawa, Naho Fujita, Yuta Kato +9
    Scientific ReportsPeer-reviewedCC BY 4.0Full textSolar & stellar physicsCrossrefDOAJ

Cite

BibTeXRIS
BibTeX
@article{sun2026mephisto,
  title = {{Mephisto: Self-improving Large Language Model–based Agents for Automated Interpretation of Multiband Galaxy Observations}},
  author = {Zechang Sun and Yuan-Sen Ting and Yaobo Liang and Nan Duan and Song Huang and Zheng Cai},
  journal = {The Astrophysical Journal Supplement Series},
  year = {2026},
  month = jul,
  doi = {10.3847/1538-4365/ae5d3a},
  url = {https://doi.org/10.3847/1538-4365/ae5d3a},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Mephisto: Self-improving Large Language Model–based Agents for Automated Interpretation of Multiband Galaxy Observations
AU  - Zechang Sun
AU  - Yuan-Sen Ting
AU  - Yaobo Liang
AU  - Nan Duan
AU  - Song Huang
AU  - Zheng Cai
PY  - 2026
DA  - 2026/07/01
JO  - The Astrophysical Journal Supplement Series
DO  - 10.3847/1538-4365/ae5d3a
UR  - https://doi.org/10.3847/1538-4365/ae5d3a
AB  - Astronomical research has long relied on human expertise to interpret complex data and formulate scientific hypotheses. In this study, we introduce Mephisto—a multiagent collaboration framework powered by large language models (LLMs) that emulates human-like reasoning to analyze multiband galaxy observations. Mephisto interfaces with the CIGALE codebase (a library of spectral energy distribution models) to iteratively refine physical models against observational data. It conducts deliberate reasoning via a tree search, accumulates knowledge through self-play, dynamically updates its knowledge base, and is validated across diverse galaxy populations—including the James Webb Space Telescope’s recently discovered “little red dot” galaxies. We show that Mephisto demonstrates the ability to infer the physical properties of galaxies from multiband photometry, positioning it as a research copilot for astronomers. Unlike prior black-box machine learning approaches in astronomy, Mephisto offers a transparent, human-aligned reasoning process that integrates with existing research practices. This work underscores the possibility of LLM-driven agent-based research for astronomy, establishes a foundation for fully automated, end-to-end artificial intelligence (AI)-powered scientific workflows, and opens up new avenues for AI-augmented discoveries in astronomy.
SN  - 0067-0049
SN  - 1538-4365
ER  -