{"id":"96a8099c-7640-487e-b8f6-81731952a5fd","arxiv_id":"2504.19962","paper_version":3,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A symbolic-regression fit, k2(M, log10 Lambda), approximates neutron-star radii from gravitational-wave-only mass and tidal-deformability measurements to within a few hundred meters for the tested equations of state.","lead":"This paper uses a machine-learning tool called symbolic regression to squeeze a short formula from simulated neutron-star data, letting astronomers estimate a neutron star's radius directly from gravitational-wave measurements of its mass and tidal deformation. If the formula holds up, it gives a fast, transparent bridge between gravitational-wave and X-ray observations of the same objects.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Non-uniqueness of the (M, Λ) → R map is real but partly acknowledged; the missing quantification of the conditional spread of R at fixed (M, Λ) leaves the 'few hundred meters' claim unverified at degeneracies.","rationale":"The reader's weakest assumption correctly identifies the non-uniqueness of the (M, Λ) → R mapping as the central risk. The paper itself demonstrates this risk in Sec. IV.B with the test1/test2 intersecting sequences and even notes that pointwise information may be insufficient, suggesting dΛ/dM as future work. My stress-test refines the concern: the reported error metrics are sequence-averaged and do not directly measure the conditional spread of R at fixed (M, Λ), so the headline accuracy of 'few hundred meters' may not hold in degeneracy regions. However, this is not an internal inconsistency; the approximation is presented as approximate and the limitation is explicitly acknowledged. The proposed binning test would settle whether the degeneracy is a rare edge case or a systematic floor. Since the reader already recommends a CONDITIONAL verdict with softened claims, my assessment does not move the verdict; it sharpens the specific check that should be required before the 'EOS-independent' language is accepted.","tokens_in":849,"tokens_out":1107,"duration_ms":56720,"concrete_test":"Run a binning analysis on the public dataset (github.com/mbejger/pysr_r-as-mlambda): for the full training+validation set of (M, Λ, R) 4-tuples, sort into small cells in (M, log10 Λ), e.g. ΔM = 0.05 M⊙ and Δlog10 Λ = 0.1; retain cells containing sequences from at least two distinct EOSs. Compute the within-cell spread of true R (max-min and 90th percentile). If a non-negligible fraction (say >5%) of multi-EOS cells have R spread exceeding 500 m, or any cell exceeds about 1 km, the pointwise formula has an intrinsic error floor not captured by the per-sequence averages, and the 'universal/EOS-independent' claims should be qualified to specify the degeneracy-limited accuracy. This test can be done with the released data and code without new simulations.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim presumes that a pointwise map (M, Λ) → R can be accurate because the conditional distribution of R at fixed (M, Λ) is narrow. The paper's own Sec. IV.B test1/test2 shows sequences from different EOSs can cross in the (M, Λ) plane, and the author explicitly states that point values of M and Λ 'may not be sufficient to recover the R value precisely' and suggests adding dΛ/dM. This is a direct admission that the map is not single-valued. The reported error metrics (ΔR, |ΔR|, |ΔR|max) are computed per sequence and then averaged over sequences; they do not measure the width of the R distribution at fixed (M, Λ). Consequently, the headline 'few hundred meters' agreement could hold on average while pointwise degeneracies produce larger errors for some EOSs. The BM165 case already reaches |ΔR|max = 452.6 m, and the constructed test shows the fit can select one branch when two are possible. Until the degeneracy is quantified over the full dataset, the 'EOS-independent' claim is stronger than the evidence supports.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses symbolic regression (PySR) on Tolman-Oppenheimer-Volkoff (TOV) solutions for piecewise-polytropic equations of state to derive a closed-form approximation for the tidal Love number k2 as a function of gravitational-wave measurable quantities M and log10 Lambda, Eq. (3). Inserting this into Eq. (2) yields an approximate neutron-star radius Rtilde(M, Lambda). The approximation is validated on a separate polytropic dataset, tested on sequences from six realistic non-polytropic EOSs, and applied to the GW170817 posterior samples, with reported average radius differences of a few hundred meters and maximum absolute differences up to 452.6 m. The paper claims the relation is EOS-independent and useful for multi-messenger radius inference.","tokens_in":17134,"tokens_out":5789,"duration_ms":58829,"significance":"If the claimed accuracy is robust, the expression would be a compact and computationally inexpensive tool for extracting radius information from gravitational-wave-only measurements, complementing existing universal relations. The paper has clear strengths: the training and validation sets are separate, the code and data are openly available, the method is reproducible, and the application to GW170817 public posteriors is a useful demonstration. The author also explicitly acknowledges a key limitation in Sec. IV.B, namely that point values of M and Lambda may not uniquely determine the radius. However, the central claim of an EOS-independent pointwise map is stronger than the evidence because the non-uniqueness of the (M, Lambda) to R relation is not quantified. The reported error metrics are per-sequence averages rather than the conditional spread of R at fixed (M, Lambda), so the worst-case behavior at degeneracies remains uncharacterized.","major_comments":[{"comment":"The intersecting-sequence test (test1/test2) demonstrates that identical or nearly identical (M, Lambda) pairs can arise from EOSs with different radii, and the text acknowledges that point values of M and Lambda may not be sufficient to recover R precisely. This non-uniqueness is load-bearing for the paper's central claim that Eqs. (2)-(3) provide an EOS-independent pointwise radius estimate. The error metrics reported in Table I and Fig. 2 (mean Delta R, mean |Delta R|, and maximum |Delta R| along each sequence) quantify per-sequence accuracy but do not measure the conditional distribution of R at fixed (M, Lambda). Please quantify this conditional spread over the full validation and realistic-EOS datasets (for example, the width of R in bins of (M, Lambda), or the error specifically at the crossing region), or explicitly qualify the claim as holding only away from such degeneracies.","section":"§IV.B, Fig. 4 and §IV.A, Table I"},{"comment":"The headline 'few hundred meters' accuracy is supported by the averaged quantities, but Table I shows |Delta R|max = 452.6 m for BM165 at M = 2.03 M_sun and a mean Delta R of +299.8 m for that entire sequence. Because radii near the maximum mass are less relevant for current gravitational-wave detections, reporting the error metrics separately for the astrophysically relevant mass window (e.g., 1.0-1.6 M_sun) would make the practical claim more precise and would clarify whether the larger discrepancies are confined to the high-mass tail of each sequence.","section":"§IV.A, Table I"}],"minor_comments":[{"comment":"The statement that Rtilde is proportional to (Lambda / log10^2(Lambda))^(1/5) and has a singularity at Lambda = 1 is misleading when applied to the full expression Eq. (3), because the +0.0137 constant term keeps k2 finite at log10 Lambda = 0. Please rephrase the sentence to refer explicitly to the leading term of the fitted expression rather than to Eqs. (2) and (3) as written.","section":"§IV.B, text after Fig. 4"},{"comment":"The inset histogram of the mass at which |Delta R| is maximal is described in the text as belonging to the validation dataset, but the caption does not state this; please add that information for clarity.","section":"§IV.A, Fig. 2 caption"},{"comment":"The reason for the mass cuts M > 0.5 M_sun and Mmax > 1.9 M_sun in the training and validation data is not explained, while Sec. II notes that the approximation is valid down to M approximately 0.25 M_sun; please clarify the relationship between these choices.","section":"§III, dataset description"},{"comment":"The caption defines Loss but only implicitly describes Score; please state explicitly that a higher Score corresponds to a better balance of accuracy versus complexity, consistent with the 'best' expression being the chosen one.","section":"Appendix A, Table A caption"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is honest about its limitations and the reproducibility practices are good. The main obstacle to acceptance is not the fit itself but the gap between the 'EOS-independent' claim and the unquantified degeneracy of the (M, Lambda) to R map. If the author adds a conditional-spread analysis at fixed (M, Lambda) and adjusts the wording of the claims accordingly, the paper would likely be acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it over the weekend. The paper does one concrete thing: it uses symbolic regression to fit k2(M, Lambda) against a large set of TOV solutions built from piecewise polytropes, then recovers R via k2 = (3/2) Lambda (GM/Rc^2)^5. The fit, Eq. 3, is compact, explicit, and apparently well-behaved; validation on realistic EOSs gives mean absolute radius errors in the tens of meters, with maxima around 450 m for BM165. Code and data are on GitHub, which counts for a lot.\n\nWhat's actually new is the explicit fitted expression, not the idea. As the reader's report says, this is a re-coordinatization of the known compactness-Love universal-relation program. But making that relation explicit, cheap, and directly applicable to (M, Lambda) posterior samples is a convenience, and the GW170817 application shows it lands in the same ballpark as the LVC's EOS-insensitive results.\n\nThe soft spots are real but not fatal. The map (M, Lambda) -> R is not single-valued; the paper's own test1/test2 intersecting sequences show two EOSs with nearly identical (M, Lambda) can have different radii. The author acknowledges this in Sec. IV.B and suggests adding dLambda/dM. The stress-test note is right that the reported error metrics are per-sequence averages and maxima, not the conditional spread of R at fixed (M, Lambda), so a 'few hundred meters' headline can coexist with larger pointwise degeneracies. Also, training and validation both come from the same piecewise-polytrope family; the realistic EOSs are a nice check but they are the exception, not the rule. The author's claim that the expression is 'largely EOS-independent' is stronger than the evidence.\n\nWhat I'd want in revision: a direct quantitative comparison with existing compactness-Love fits, and a plot or quantify the conditional distribution of R at fixed (M, Lambda) across the validation set. That would turn a good empirical note into a referenceable tool.\n\nBottom line: it deserves a serious referee, with a request to soften 'EOS-independent' and add the baseline comparison. I'd cite it if I were doing GW radius inference; it's a handy shortcut. Reading group? Maybe, as a short methods talk.","headline":"A useful, honest symbolic-regression fit for R(M, Lambda) with real soft spots around non-uniqueness; deserves refereeing with softened claims and a baseline comparison.","tokens_in":17691,"tokens_out":2789,"would_cite":true,"duration_ms":28435,"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":"A single symbolic expression approximates the neutron-star radius from only mass and tidal deformability, recovering radii to within a few hundred meters across a broad range of equations of state.","keywords":["neutron stars","gravitational waves","tidal deformability","tidal Love number","symbolic regression","equation of state","radius inference","GW170817"],"falsifier":"Evaluate Eqs. (2)–(3) on stellar-structure sequences for equations of state with strong first-order phase transitions over the mass range $1$–$2\\,M_\\odot$; if any sequence yields $|\\Delta R|>0.5$ km at a mass where $\\Lambda$ is well constrained, the claimed few-hundred-meter accuracy fails for that class of equations of state. An alternative is to find two realistic equations of state whose $(M,\\Lambda)$ curves cross at a point where their radii differ by more than the reported error.","tokens_in":16679,"feed_emoji":"🌌","tokens_out":10976,"duration_ms":95301,"temperature":0.7,"pith_summary":"This paper claims that a single compact formula, discovered by symbolic regression, can recover a neutron star's radius from only two gravitational-wave observables: the component mass $M$ and tidal deformability $\\Lambda$. If true, gravitational-wave detectors could estimate radii directly, without requiring a simultaneous electromagnetic measurement or a full equation-of-state reconstruction. The formula approximates the tidal Love number $k_2$ as a function of $M$ and $\\Lambda$, then inverts the standard $k_2$–$\\Lambda$–$R$ relation to obtain $R$. The author reports average errors of a few hundred meters on validation polytropes and on several realistic non-polytropic equations of state, and applying the formula to the GW170817 event yields radii consistent with previously reported inferences.","feed_headline":"One equation maps gravitational waves to neutron-star radii","feed_subtitle":"A symbolic expression turns mass and tidal-deformability measurements into few-hundred-meter radius estimates.","key_machinery":"The load-bearing object is the dimensionless tidal Love number $k_2$, defined through $k_2=\\frac{3}{2}\\Lambda\\left(\\frac{GM}{Rc^2}\\right)^5$, which connects the measured deformability $\\Lambda$ to the radius $R$. The paper uses a symbolic-regression search to find a simple closed-form surrogate $\\tilde{k}_2(M,\\Lambda)$ — Eq. (3) — by fitting stellar-structure data. Substituting $\\tilde{k}_2$ into the inverted Love-number relation gives $\\tilde{R}(M,\\Lambda)$, turning radius estimation into the evaluation of one explicit algebraic expression and bypassing any need to assume a specific equation of state or solve the stellar-structure equations anew.","core_discovery":"The paper's central claim is that, to good approximation, the radius $R$ of a non-rotating neutron star is a function of only $(M,\\Lambda)$, the two quantities directly measured during a binary inspiral. Concretely, it proposes $\\tilde{k}_2(M,\\Lambda)=0.0351(\\log_{10}\\Lambda)^2\\left(\\frac{M}{M+2.05}-0.109\\right)+0.0137$ for the tidal Love number, with $M$ in solar masses, and then $\\tilde{R}(M,\\Lambda)=\\left(\\frac{3\\Lambda}{2\\tilde{k}_2}\\right)^{1/5}\\frac{GM}{c^2}$. The relation is trained on solutions to the relativistic stellar-structure equations for piecewise polytropic equations of state and validated on both unseen polytropes and realistic dense-matter theories, with typical absolute radius differences of a few hundred meters. Applied to posterior samples from GW170817, it produces radii that fall inside the 50% and 90% credible regions of the published comparison distributions.","pith_inferences":["Augmenting the pointwise formula with a local slope, such as $d\\Lambda/dM$, would likely resolve the intersection degeneracy the paper identifies, though it would give up the pure point-estimate character.","Because the fitted relation is cheap and explicit, it could seed hierarchical Bayesian analyses as an informed prior or act as a fast likelihood surrogate in parameter-estimation pipelines.","Retraining the same symbolic-regression pipeline on data that include hybrid stars or strong first-order phase transitions could extend the formula to the regimes where the current version shows its largest errors near the maximum mass.","A direct test against an independent electromagnetic radius measurement from X-ray pulse timing would be a natural next step; the paper does not include such a comparison."],"forward_implications":["Gravitational-wave-only radius estimates become a one-line evaluation, so individual binary-neutron-star events can yield radius measurements without costly equation-of-state sampling or structure integrations.","For the mass range roughly $1$–$2.5\\,M_\\odot$, the reported few-hundred-meter accuracy is competitive with typical equation-of-state-based inference, making the formula useful as a fast cross-check or as a summary statistic.","The expression extrapolates to non-polytropic equations of state and to sequences partially outside the training region, indicating that coverage of the physical equation-of-state space, rather than the specific training family, is the main driver of accuracy.","Applied to GW170817, the formula reproduces the published radius posterior distributions within the reported credible regions, so the relation can serve as a straightforward consistency test for future detections."],"supporting_citations":[{"why":"Provides the GW170817 detection whose posterior samples are used to test the radius formula on real gravitational-wave data.","marker":"[1]"},{"why":"Supplies the training and validation datasets of piecewise-polytropic stellar-structure sequences.","marker":"[17]"},{"why":"Provides the relativistic stellar-structure equations whose solutions produce the mass-radius-deformability data.","marker":"[33]"},{"why":"Companion paper defining the structure equations used for computing neutron-star models.","marker":"[34]"},{"why":"Establishes the tidal Love number k2 and its relation to the tidal deformability used in the inversion.","marker":"[36]"},{"why":"Supplies the symbolic-regression implementation used to discover the expression for k2.","marker":"[38]"},{"why":"Reports the GW170817 radius posterior distributions used for comparison.","marker":"[52]"}],"fun_headline_variants":["Gravitational waves alone yield neutron-star radii","Symbolic regression maps GW data to NS radii","One formula turns chirps into neutron-star sizes","Radius from mass and tidal deformability alone"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result relies on the assumption that one EOS-independent, single-valued function $k_2(M,\\Lambda)$ exists, so that every $(M,\\Lambda)$ pair corresponds to a unique radius; the paper's own test with intersecting sequences shows that two different equations of state can share the same mass and deformability while having different radii, giving the formula an intrinsic accuracy floor.","fun_headline_variants_meta":{"raw":{"variants":["Gravitational waves alone yield neutron-star radii","Symbolic regression maps GW data to NS radii","One formula turns chirps into neutron-star sizes","Radius from mass and tidal deformability alone"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000195,"raw_usage":{"total_tokens":1365,"prompt_tokens":964,"completion_tokens":401,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":580,"completion_tokens_details":{"reasoning_tokens":343}},"tokens_in":580,"tokens_out":401,"duration_ms":3791,"temperature":1.0,"reasoning_tokens":343,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:38:42.985648+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate Eqs. (2)–(3) on stellar-structure sequences for equations of state with strong first-order phase transitions over the mass range $1$–$2\\,M_\\odot$; if any sequence yields $|\\Delta R|>0.5$ km at a mass where $\\Lambda$ is well constrained, the claimed few-hundred-meter accuracy fails for that class of equations of state. An alternative is to find two realistic equations of state whose $(M,\\Lambda)$ curves cross at a point where their radii differ by more than the reported error.","supporting_citations":[{"cited_title":"Machine Learning-Based Analytical Expressions for Gray-Body Factors and Application to Primordial Black Holes","cited_arxiv_id":"2504.18270","evidence_quote":"Provides the relativistic stellar-structure equations whose solutions produce the mass-radius-deformability data."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Companion paper defining the structure equations used for computing neutron-star models."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the symbolic-regression implementation used to discover the expression for k2."},{"cited_title":"3 presented the recovered radii for both BNS components, using various assumptions about the EOS; see LIGO-P1800115 for the data behind the figure","cited_arxiv_id":null,"evidence_quote":"Reports the GW170817 radius posterior distributions used for comparison."}],"review_version":1}