{"id":"2968cba3-a77c-48f1-a62f-c66273489cba","arxiv_id":"2508.19297","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"STCTM is a Bayesian, parallelized retrieval framework that models stellar contamination (TLSE) in exoplanet transmission spectra and enables retrievals on out-of-transit stellar spectra.","lead":"This paper presents STCTM, an open-source Python framework for modeling stellar contamination (the transit light source effect) in exoplanet transmission spectra, including retrievals on out-of-transit stellar spectra. It addresses a key systematic effect that can mimic or mask planetary atmospheric features in JWST observations of small planets around M dwarfs.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"STCTM's inference claim rests on an unvalidated TLSE forward model; with no equation-level description or injection-recovery benchmark, grid/systematic errors could bias the retrieved 'compatible' stellar parameters.","rationale":"The reader's weakest assumption and my concern coincide: the correctness of the TLSE forward model, including its reliance on precomputed stellar grids and simplified spot/facula parametrization, is the condition on which the central inference claim depends. I agree with the CONDITIONAL verdict because the paper does not provide validation experiments, a code repository link, or equation-level details in the submitted text. The concern is not that the framework is wrong, but that the central claim is not yet independently checkable from the manuscript. Thus I recommend leaving the verdict unchanged: the paper should be accepted only once the repository, documentation, and an explicit validation test confirm that the forward model retrieves known inputs without bias.","tokens_in":5678,"tokens_out":3659,"duration_ms":52841,"concrete_test":"Run an injection-recovery benchmark: generate synthetic JWST/NIRISS transmission spectra from a known spot/facula configuration (e.g., T_spot=2800 K, f_spot=0.1, f_fac=0.2) using an independent forward model, then retrieve with STCTM using the same default grids. Repeat for multiple grid choices (PHOENIX vs SPHINX) and for 10 random parameter draws. If the 68% credible intervals for T_spot, f_spot, f_fac exclude the truth, or if the recovered contamination spectrum deviates from the injected one by more than the assumed noise, the claim that retrieved parameters are 'compatible with the observations' is not yet supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that STCTM can retrieve the range of stellar surface parameters compatible with a transmission spectrum when no planetary contribution is present. This requires the forward model to be accurate enough for high-S/N JWST data. The paper gives no equation-level specification of that forward model and no validation experiment. The forward model rests on two externally inherited assumptions: (i) each heterogeneity is represented by a single-temperature region using precomputed PHOENIX/SPHINX spectra, and (ii) the contrast spectrum is insensitive to limb-darkening, magnetic activity, and spot/facula temperature structure. Both are known approximations for M dwarfs. If the adopted stellar grid is systematically off in T_eff, [Fe/H], or spot temperature, the inferred contamination spectrum is biased in a wavelength-dependent way, and the conclusion that observations are compatible with no planetary atmosphere can be wrong. The cited applications (Lim et al. 2023; Radica et al. 2025) use this exact inference to set atmospheric limits, so this is the load-bearing point. The code may well be correct; the concern is that this paper does not yet provide the evidence needed to check it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"STCTM is an open-source Bayesian retrieval framework designed to model the transit light source effect (TLSE) in exoplanet transmission spectra and to infer the stellar surface parameters that are consistent with observations when no planetary signal is assumed. The paper describes the high-level workflow (configuration via input.toml, MCMC fitting with emcee, post-processing and publication-ready output), highlights the exotune sub-module for retrieving stellar parameters from out-of-transit spectra, and lists several recent peer-reviewed applications (Lim et al. 2023; Radica et al. 2025; Piaulet-Ghorayeb et al. 2024, 2025; Ahrer et al. 2025; Roy et al. 2023). The paper does not, however, provide an equation-level specification of the forward model, any validation benchmarks, or quantitative tests of the retrieval pipeline.","tokens_in":5979,"tokens_out":4080,"duration_ms":49095,"significance":"If the framework performs as claimed, it fills a clear community need: a flexible, modular, open-source tool for TLSE modeling and inference that supports arbitrary stellar model grids and is parallelized for tractable MCMC retrievals. The existing literature applications are a notable strength, providing indirect evidence of practical utility and community adoption. The emphasis on user-configurable inputs, reproducible outputs, and integration with widely-used stellar model grids (PHOENIX, SPHINX, MSG) is also valuable. The central scientific claim, however, rests on the correctness of the underlying forward model and the inference procedure. Because the paper itself contains no derivation or test of that model, the evidence currently available to a reader is mostly circumstantial; the paper's credibility would be substantially improved by explicitly presenting the forward-model equations and a validation demonstration.","major_comments":[{"comment":"The paper does not specify the forward model used to compute the TLSE contrast spectrum. It never defines the relationship between spot/faculae covering fractions, temperature contrasts, and the observed transmission spectrum. Without this specification, readers cannot assess whether the MCMC samples the intended parameter space or whether the implementation is correct. Please add the relevant equations (e.g., the standard contamination factor expression relating observed and true transit depth to spot/faculae properties), or at minimum give an explicit pointer to the equations used in Lim et al. 2023 / Radica et al. 2025 and the online documentation, and state how the stellar model grids are interpolated (e.g., via MSG) and how synthetic spectra for spot/faculae regions are combined.","section":"Summary / Main features of the code"},{"comment":"No validation benchmarks or synthetic tests are presented in the paper. The 'Uses of STCTM in the literature' section provides examples of application, but a software paper whose central claim is a reliable inference framework needs more direct evidence. Please include at least one injection-recovery test showing that STCTM recovers input spot/faculae parameters and correctly identifies 'no planetary contribution' when none is present. Alternatively, include a representative retrieval on a published dataset and compare with the published posteriors. The test should be described in the paper itself (not only linked in the documentation) so that the reader can verify the pipeline's behavior.","section":"Documentation / Uses of STCTM in the literature"},{"comment":"The retrieved parameters are sensitive to systematic errors in the adopted stellar model grids and to the simplified parameterization of stellar heterogeneities as single-temperature spot/faculae regions with covering fractions. The paper mentions an error-inflation parameter only for exotune retrievals, not for the transmission-spectrum TLSE retrievals. Since the central claim is that STCTM can infer the range of stellar parameters 'compatible with the observations in the absence of any planetary contribution,' systematic model error could directly bias this range. Please add a discussion of this limitation and, ideally, demonstrate robustness by running a retrieval with two different model grids (e.g., PHOENIX vs. SPHINX) or by including an error-inflation / model-marginalization term in the transmission-spectrum case as well.","section":"Main features / exotune"}],"minor_comments":[{"comment":"The Piaulet-Ghorayeb et al. 2025 reference has a URL pointing to a 2024 ApJ paper (2024ApJ...974L..10P). Please correct or clarify whether this is the intended citation.","section":"References"},{"comment":"The phrase 'spectral contrasts between bright and dark spots' is a little ambiguous: spots are usually the dark regions and faculae the bright ones. Consider rephrasing to 'spots and faculae' for clarity.","section":"Summary"},{"comment":"For POSEIDON, the description 'including TLSE-only retrievals on transmission spectra' would benefit from a precise reference to the relevant POSEIDON documentation or paper, since the current sentence is vague.","section":"Similar Tools"},{"comment":"The version number (v2.1.1) is mentioned, but no repository URL appears in the printed text. For reproducibility, please include the repository and documentation URLs explicitly in the manuscript body or a footnote.","section":"Future Developments"}],"recommendation":"major_revision","confidential_remarks":"The paper is quite short and reads as a JOSS-style software description. For a more general journal, the lack of any equation-level forward-model specification and validation experiment is the main blocker. The existing literature applications are encouraging, but they do not substitute for a direct test in this manuscript. A major revision that adds the governing equations and at least one synthetic retrieval benchmark would make the central claim checkable and would, in my view, render the paper suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nSTCTM is a genuinely useful piece of software: it wraps TLSE retrieval on transmission spectra and out-of-transit stellar retrieval (exotune) into one parallelized, open-source Bayesian framework with pluggable stellar grids. That combination is real and addresses a community need. The paper describes the architecture clearly and points to published applications (Lim et al. 2023, Radica et al. 2025) that used the code, which is meaningful evidence that it works in practice.\n\nWhat is new here is modest but not zero: existing tools do one of the two retrievals but not both, and STCTM claims flexibility in input formats and model grids plus parallelization. The writing is honest about the TLSE model deriving from Rackham et al. (2018) and about the code's reliance on emcee and external spectral grids.\n\nThe soft spots are real, though not disqualifying. The paper gives no equation-level description of the forward model and no injection-recovery or validation benchmark in the text; a skeptical reader cannot check the correctness of the TLSE implementation from the paper alone. That matters because the code is being used to set atmospheric limits, and grid systematics or implementation bugs would propagate directly into those conclusions. The statement of need also overstates the gap: POSEIDON already does TLSE retrievals on transmission spectra, and StellarFit does out-of-transit retrievals. The paper acknowledges both in the 'Similar Tools' section, but the earlier rhetoric could be tighter.\n\nThat said, this is a JOSS-style software paper, not a methods paper. The full proof of the pudding is in the repository and docs, and the paper points to them. The stress-test concern about forward-model accuracy is a real caveat that applies to the entire TLSE modeling enterprise, not a flaw unique to STCTM. I'd want to see a test suite or a cross-check against POSEIDON in the repo before fully trusting the numbers, but that's what a referee should ask for, not a reason to desk-reject.\n\nVerdict: this deserves peer review. The tool is used in the literature and fills a niche. For a reading group, it's more of a 'maybe'—useful context for interpreting JWST transmission spectrum papers, but not a deep conceptual read. I'd cite it if I'm working on M-dwarf transmission spectra, since it's becoming the standard tool in that niche.\n\nRecommendation: send to referees, but ask for a validation figure or a link to a reproducible benchmark in the next revision.","headline":"Useful and already-used tool paper that combines TLSE and stellar retrievals, but it lacks an in-paper validation benchmark and slightly overstates the gap against POSEIDON and StellarFit.","tokens_in":6439,"tokens_out":3097,"would_cite":true,"duration_ms":36395,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper presents STCTM, a Bayesian framework that models how star spots and faculae contaminate exoplanet transmission spectra, letting observers test whether an apparent atmospheric signal could actually be stellar contamination.","keywords":["transit light source effect","stellar contamination","exoplanet transmission spectroscopy","M dwarf stars","star spots","faculae","Bayesian retrieval","MCMC"],"falsifier":"Take a star with an independently measured surface map, for example from Doppler imaging, and a JWST transmission spectrum of one of its planets; run STCTM and compare the retrieved spot covering fractions and temperature contrasts to the mapped values. A systematic mismatch would show that the forward model's simplified surface representation is insufficient for real stellar surfaces.","tokens_in":5587,"feed_emoji":"🪐","tokens_out":6145,"duration_ms":69436,"temperature":0.7,"pith_summary":"This paper introduces STCTM, a flexible, open-source Bayesian retrieval framework for the transit light source effect (TLSE). The TLSE is the wavelength-dependent imprint that unocculted stellar spots and faculae leave on a transmission spectrum, and it can mimic or mask planetary atmospheric features, especially for small planets around M dwarfs. The paper argues that the community needs a tool that can model this contamination generically, fit it to any transmission spectrum, and infer the stellar surface parameters that could explain the data without invoking a planetary atmosphere. STCTM also includes an exotune sub-module that retrieves stellar parameters directly from out-of-transit spectra, providing data-driven priors on how much contamination a host star can inject. If the framework works as claimed, it gives observers a practical way to test the robustness of exoplanet atmosphere detections.","feed_headline":"New code tells star spots from exoplanet atmosphere signals","feed_subtitle":"Open-source STCTM fits spots and faculae to JWST/HST spectra, guarding against false planetary-atmosphere claims.","key_machinery":"The load-bearing object is the TLSE forward model: a parameterized stellar surface made of spots and faculae with covering fractions and temperature contrasts, combined with precomputed synthetic spectra from model grids such as PHOENIX or SPHINX. The model produces a wavelength-dependent contamination spectrum that MCMC compares to the observed transmission spectrum. Exotune applies the same surface parameterization to out-of-transit spectra. The framework's genericity comes from allowing any user-specified stellar model grid and input data format.","core_discovery":"The central claim is that the transit light source effect can be forward-modeled and retrieved with a general, user-tunable Bayesian framework. STCTM parameterizes the stellar surface using covering fractions and temperature contrasts for spots and faculae, computes the resulting contamination spectrum from user-supplied synthetic stellar model grids, and fits this model to observed transmission spectra with MCMC. In exotune mode, the same machinery is applied to out-of-transit stellar spectra, allowing independent constraints on the contamination a host star can produce. The paper positions this as filling a gap: existing tools are either not public, not built for inference, serial-only, or","pith_inferences":["The same spot-and-faculae retrieval applied to out-of-transit spectra could be time-resolved to track stellar activity evolution, an extension the paper does not explicitly develop.","A natural next step is adding wavelength-dependent limb darkening or temperature gradients across spots and faculae; the current simplified surface model may absorb such effects into broader parameter uncertainties.","If STCTM's TLSE-only fits are statistically preferred for a sizable sample of M-dwarf planets, estimates of how common featureless or atmosphere-free rocky planets are may need to be revisited; the paper enables such an analysis but does not quantify it."],"forward_implications":["Transmission spectra of small planets around M dwarfs can be checked against a TLSE-only model; if that model fits, an apparent atmospheric detection is not secure without further evidence.","Out-of-transit stellar spectra can be used to place data-driven priors on the amplitude and wavelength shape of stellar contamination for planets around the same host star.","Because stellar model grids are plug-in, results can be cross-checked across different synthetic spectral libraries without rewriting the retrieval code.","Fully parallelized MCMC makes the inference fast enough for routine application to JWST and HST transmission spectra.","The framework outputs model comparison metrics, sample spectra, and parameter samples, making contamination fits reproducible across different studies."],"supporting_citations":[{"why":"Defines the transit light source effect and shows how it produces false spectral features in M-dwarf transmission spectra, establishing the problem STCTM addresses.","marker":"(Rackham et al., 2018)"},{"why":"Identifies stellar contamination as a limiting factor in the atmospheric characterization of terrestrial exoplanets, motivating the need for robust retrieval tools.","marker":"(TRAPPIST-1 JWST Community Initiative et al., 2024)"},{"why":"Supplies the PHOENIX synthetic stellar atmosphere grid, one of the model grids STCTM uses to compute spot and faculae spectra.","marker":"(Husser et al., 2013)"},{"why":"Supplies the SPHINX M-dwarf spectral grid, providing synthetic spectra tailored to the cool stars most affected by stellar contamination.","marker":"(Iyer et al., 2023)"},{"why":"Provides the emcee MCMC sampler that powers the Bayesian retrievals in STCTM.","marker":"(Foreman-Mackey et al., 2013)"},{"why":"Represents an existing atmospheric retrieval tool with TLSE-only capabilities, serving as a comparison point for STCTM's transmission-spectrum retrievals.","marker":"(MacDonald & Madhusudhan, 2024)"},{"why":"Provides StellarFit, a comparison tool for retrievals on out-of-transit stellar spectra, situating the exotune sub-module.","marker":"(Radica et al., 2025)"},{"why":"Provides the Eureka! time-series pipeline whose stage-3 output exotune can take as input, supporting the out-of-transit retrieval workflow.","marker":"(Bell et al., 2022)"}],"fun_headline_variants":["New tool separates star spots from exoplanet atmosphere signals","Bayesian code untangles stellar spots from exoplanet spectra","STCTM disentangles star spots from exoplanet atmospheric signals","Code models star spots to reveal true exoplanet atmospheres","Separate stellar contamination from exoplanet atmosphere with STCTM"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The load-bearing premise is that a few uniform spots and faculae with fixed temperature contrasts, rendered with precomputed stellar model grids, faithfully represent the real heterogeneous surface of the star; if the grids are biased or the parameterization misses key effects, the retrieved contamination spectrum and any conclusion that no planetary atmosphere is required will be biased.","fun_headline_variants_meta":{"raw":{"variants":["New tool separates star spots from exoplanet atmosphere signals","Bayesian code untangles stellar spots from exoplanet spectra","STCTM disentangles star spots from exoplanet atmospheric signals","Code models star spots to reveal true exoplanet atmospheres","Separate stellar contamination from exoplanet atmosphere with STCTM"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000209,"raw_usage":{"total_tokens":1266,"prompt_tokens":788,"completion_tokens":478,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":532,"completion_tokens_details":{"reasoning_tokens":405}},"tokens_in":532,"tokens_out":478,"duration_ms":5822,"temperature":1.0,"reasoning_tokens":405,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T16:26:26.530918+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a star with an independently measured surface map, for example from Doppler imaging, and a JWST transmission spectrum of one of its planets; run STCTM and compare the retrieved spot covering fractions and temperature contrasts to the mapped values. A systematic mismatch would show that the forward model's simplified surface representation is insufficient for real stellar surfaces.","supporting_citations":[],"review_version":1}