An LLM-based multi-agent system (FORMA) verifies DESI spectral classifications with 95.5% agreement to expert adjudication by reconstructing expert reasoning into an auditable workflow.
https://arxiv.org/abs/2510.17960
13 Pith papers cite this work. Polarity classification is still indexing.
13
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
years
2026 13roles
background 2polarities
background 2representative citing papers
A spiked signal-plus-noise model yields separation ratios that partition multimodal problems into four regimes where alignment, prediction, both, or neither succeed.
A single end-to-end Transformer model unifies stellar labels from heterogeneous spectroscopic surveys into a self-consistent scale without post-hoc recalibration.
Vision-language models underperform specialized astronomical methods on real observational data, with accuracy improving when physical explanations are provided in prompts and when raw numerical measurements replace rendered plots.
A masked autoencoder model trained on 4.7 million DESI single-fiber observations predicts emission-line maps from images that match independent MaNGA IFU data without any IFU training.
Two-stage LLM framework infers stellar parameters and ~20 elemental abundances from spectra, showing performance gains with increasing data volume.
PITA, a new semi-supervised deep learning algorithm, outperforms prior photo-z methods by using a triple-task loss on images, colors, and available redshifts to produce a smooth latent space.
ORACLE-2 multimodal classifiers raise macro F1 from 0.52-0.66 (light-curve only) to 0.73 on ZTF Bright Transient Survey data and reach 0.88 on simulated ELAsTiCC data.
Benchmark of Affine, AIM, JetFormer and VQ-VAE tokenizers on galaxy images shows decoupled reconstruction and representation performance with no consistent winner.
Pretraining data composition can be used to engineer neural scaling laws in hadronic jet classification toward data-heavy rather than model-size-heavy regimes.
Conformal methods achieve near-nominal 90% coverage on galaxy property regression with AION-1 embeddings while LVD additionally delivers local validity, outperforming Deep Ensembles and MC Dropout.
A multimodal neural network trained on MPA-JHU references produces SFR, stellar mass, and metallicity estimates for 547 million low-redshift galaxies in DESI LS DR10.
AI techniques for photometric redshift estimation have converged and are now limited by the size, systematics, and selection effects in spectroscopic training samples rather than by methodology.
citing papers explorer
-
Executable verification through formalized expert reasoning in astronomical spectroscopy
An LLM-based multi-agent system (FORMA) verifies DESI spectral classifications with 95.5% agreement to expert adjudication by reconstructing expert reasoning into an auditable workflow.
-
When to Align, When to Predict: A Phase Diagram for Multimodal Learning
A spiked signal-plus-noise model yields separation ratios that partition multimodal problems into four regimes where alignment, prediction, both, or neither succeed.
-
Homogeneous Stellar Parameters from Heterogeneous Spectra with Deep Learning
A single end-to-end Transformer model unifies stellar labels from heterogeneous spectroscopic surveys into a self-consistent scale without post-hoc recalibration.
-
A systematic evaluation of vision-language models for observational astronomical reasoning tasks
Vision-language models underperform specialized astronomical methods on real observational data, with accuracy improving when physical explanations are provided in prompts and when raw numerical measurements replace rendered plots.
-
Integral Field Unit Spectroscopy with One Fiber
A masked autoencoder model trained on 4.7 million DESI single-fiber observations predicts emission-line maps from images that match independent MaNGA IFU data without any IFU training.
-
Spectra as Language: Large Language Models for Scalable Stellar Parameter and Abundance Inference
Two-stage LLM framework infers stellar parameters and ~20 elemental abundances from spectra, showing performance gains with increasing data volume.
-
Optimizing Deep Learning Photometric Redshifts for the Roman Space Telescope with HST/CANDELS
PITA, a new semi-supervised deep learning algorithm, outperforms prior photo-z methods by using a triple-task loss on images, colors, and available redshifts to produce a smooth latent space.
-
Leveraging Multimodality for Real-Time Classification of Transients and Variables found by the Zwicky Transient Facility
ORACLE-2 multimodal classifiers raise macro F1 from 0.52-0.66 (light-curve only) to 0.73 on ZTF Bright Transient Survey data and reach 0.88 on simulated ELAsTiCC data.
-
The Galaxy's Guide to the Tokenizer: A Benchmark for Scientific Foundation Models
Benchmark of Affine, AIM, JetFormer and VQ-VAE tokenizers on galaxy images shows decoupled reconstruction and representation performance with no consistent winner.
-
Towards Engineering Scaling Laws with Pretraining Data Composition
Pretraining data composition can be used to engineer neural scaling laws in hadronic jet classification toward data-heavy rather than model-size-heavy regimes.
-
Beyond Point Estimates: Benchmarking Uncertainty Quantification Methods on the AION-1 Astronomical Foundation Model
Conformal methods achieve near-nominal 90% coverage on galaxy property regression with AION-1 embeddings while LVD additionally delivers local validity, outperforming Deep Ensembles and MC Dropout.
-
A Value-added Physical Properties Catalog for Low-redshift Galaxies from DESI Legacy Imaging Surveys DR10
A multimodal neural network trained on MPA-JHU references produces SFR, stellar mass, and metallicity estimates for 547 million low-redshift galaxies in DESI LS DR10.
-
Machine Learning Techniques for Astrophysics and Cosmology: Photometric Redshifts
AI techniques for photometric redshift estimation have converged and are now limited by the size, systematics, and selection effects in spectroscopic training samples rather than by methodology.