A Missing Latent, Not a Missing Simulator: Radius-Augmented Inference for Real JWST Retrieval

Angshuman Chakravertty; P V V Raj

astro-ph.IMSep 30, 2026

A Missing Latent, Not a Missing Simulator: Radius-Augmented Inference for Real JWST Retrieval

Authors: Angshuman Chakravertty, P V V Raj

Organizations: SVKM’s NMIMS (Deemed to be University)

Abstract

Amortized simulation-based inference (SBI), which is trained on radiative-transfer simulators, recovers exoplanet atmospheres accurately on synthetic James Webb Space Telescope (JWST) spectra but collapses when it comes to real reduced spectra. The flow posterior collapsed on real WASP-39b (importance-sampling effective sample size (ESS) = 1, best-fit \c{hi}2/N = 301), and such a failure is usually blamed on missing the forward-model physics, but ruling these levers out with nested sampling first (a temperature gradient, SO2 opacity, and a high-fidelity opacity set) leaves the fit unchanged, meaning the collapse is not from the simulator misspecification but instead from a missing latent, the planet radius. To fix this, we build MIRAGE, a radius-augmented flow-matching posterior calibrated against an independent nested-sampling reference with importance sampling and an optimal-transport map, which yields a physical and literature-consistent retrieval of real WASP-39b (with \c{hi}2/N from 301 to 0.06). This same method transfers unchanged across two instruments and three real JWST targets, including one spectrum self-reduced end-to-end from raw Mikulski Archive for Space Telescopes(MAST) data. The lesson is cross-domain, as a latent the simulator encodes but the inference omits can masquerade as misspecification.

Figures & tables

Figure 1: MIRAGE architecture. Stages 1 and 2 (simulation and amortized training) run offline, and Stage 3 retrieves on a real JWST spectrum per target and calibrates the flow posterior against an independent nested-sampling reference.

Forward-model variant (radius fixed unless noted)	best-fit $T$ (K)	$\chi^{2}/N$
Physical atmosphere, isothermal	700	38.44
+ temperature gradient	700	38.48
+ SO 2 opacity	700	38.46
Free planet radius	606	0.76

Table 1: Ruling out forward-model fidelity, nested sampling first. Each variant is fit to real WASP-39b with the radius held fixed except the last row and

\chi^{2}/N

is measured over the 47 real-covered bins.

Figure 2: Multi-target retrievals. The same radius-augmented, calibrated method applied unchanged to three real JWST targets that span planet class, instrument, and data provenance. The top row shows the MIRAGE posterior against the real spectrum, and the bottom row shows the recovered planet-radius posterior against the nested-sampling anchor and the literature value.

Planet	Instrument	Radius (truth) $R_{J}$	Coverage vs NS
WASP-39b	NIRSpec PRISM	1.227 (1.28)	7/7
WASP-96b	NIRISS SOSS	1.196 (1.20)	7/7
K2-18b	NIRISS SOSS	0.232 (0.24)	6/7

Table 2: The same radius-augmented, calibrated method applied unchanged to three real JWST targets. Recovered planet radius, with the literature value in parentheses, and the coverage of the nested-sampling reference values inside the calibrated credible intervals, taken over the seven inferred parameters.

Simulation-Based Inference

Simulation-based inference estimates parameters and uncertainty from simulators when evaluating a tractable likelihood is difficult or impossible.

See all 87 papers

Information-Preserving Domain Transfer with Unlabeled Data in Misspecified Simulation-Based Inference
May 7, 2026Joon Jang, Eunho Jeong, Kyu Sung Choi +1Synthetic-to-Real Domain Adaptation Simulation-Based Inference
Department of Biomedical Sciences, Seoul National University, Seoul, Republic of Korea · Department of Applied Bioengineering, Graduate School of Convergence Science and Technology, Seoul National University, Seoul, Republic of Korea · Department of Radiology, Seoul National University Hospital, Seoul, Republic of Korea +3
Efficient reduction of stellar contamination and noise in planetary transmission spectra using neural networks
Feb 10, 2026David S. Duque-Castaño, Lauren Flor-Torres, Jorge I. ZuluagaDenoising Autoencoders Autoencoders
SEAP/FACom, Instituto de Física - FCEN, Universidad de Antioquia, Calle 70 No. 52-21, Medellín, Colombia
SIMBA: ABidirectional Retrieval Forward Simulation Framework for Modeling FY-4A GIIRS Hyperspectral Infrared Radiances Toward NWP Applications
Jun 18, 2026Jingdong Shen, Fu Wang*, Qifeng Lu +4Remote Sensing Hyperspectral Imaging
School of Computer Science and Engineering, Sichuan University of Science & Engineering, Yibin 644000, China · State Key Laboratory of Severe Weather Meteorological Science and Technology, CMA Earth System Modeling and Prediction Centre and Key Laboratory of Earth System Modeling and Prediction, China Meteorological Administration, Beijing, 100081

May 7, 2026J/K move · Enter open · S save

Information-Preserving Domain Transfer with Unlabeled Data in Misspecified Simulation-Based Inference

Joon Jang, Eunho Jeong, Kyu Sung Choi, Hyeonjin Kim

Department of Biomedical Sciences, Seoul National University, Seoul, Republic of Korea · Department of Applied Bioengineering, Graduate School of Convergence Science and Technology, Seoul National University, Seoul, Republic of Korea · Department of Radiology, Seoul National University Hospital, Seoul, Republic of Korea+3

Synthetic-to-Real Domain Adaptation Simulation-Based Inference Unsupervised Domain Adaptation Domain Adaptation Bayesian Inference Model Misspecification

Open paper AskPDF arXiv

Simulation-based inference (SBI) provides amortized Bayesian parameter inference from simulator-generated data without requiring explicit likelihood evaluation. Its reliability can degrade under model misspecification, where real-world observations are not well represented by the simulator used for training. Existing methods using unlabeled real-world data often align simulated and real-world data distributions, but marginal alignment alone does not directly preserve parameter-relevant information needed for posterior inference. We propose SPIN, an SBI framework with parameter-relevant information-preserving domain transfer using unlabeled, unpaired real-world observations. During training, SPIN translates labeled simulator observations toward the real-world domain and back to the simulator domain, using the original simulator labels to encourage domain transfer that preserves parameter-relevant mutual information. At test time, the learned real-to-simulator transport maps real-world observations into the simulator domain for posterior inference, without requiring real-world parameter labels or paired real--simulator observations. Across controlled synthetic and physical real-world benchmarks, SPIN improves real-world posterior inference, with the improvement becoming clearer as misspecification increases.

Abstract

Simulation-Based Inference

Explore similar work

Information-Preserving Domain Transfer with Unlabeled Data in Misspecified Simulation-Based Inference

Efficient reduction of stellar contamination and noise in planetary transmission spectra using neural networks

SIMBA: ABidirectional Retrieval Forward Simulation Framework for Modeling FY-4A GIIRS Hyperspectral Infrared Radiances Toward NWP Applications