{"id":"4e385179-b987-4712-b995-cfd8dfc78395","arxiv_id":"2602.21733","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Combining BOSS power spectrum and bispectrum with DESI and DES lensing data yields a sound-horizon-free H0 = 70.2 ± 2.3 km/s/Mpc, with a 1.8σ BAO scale deviation that is scale-cut dependent.","lead":"Using galaxy clustering, weak lensing, and CMB lensing data without relying on the sound horizon, the authors measure the Hubble constant H0 ≈ 70.2 km/s/Mpc at 3–4% precision. The result sits between early- and late-universe values, and a small BAO-scale anomaly hints at evolving dark energy.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Residual sound-horizon information via the baryon-suppression broadband is asserted negligible only for power-spectrum analyses; the new BOSS bispectrum at k_Bmax=0.23 is where the method becomes scale-cut dependent, so the sound-horizon-free label is not yet secured.","rationale":"The reader's weakest_assumption—residual sound-horizon information through β r_s—is indeed the load-bearing concern for the paper's central 'sound-horizon-free' claim. I agree with the reader that this is the least secure step, and I sharpen it by pointing to the bispectrum extension: Ref. [19]'s negligibility result was derived for power-spectrum-only analyses, but this paper uses the same α_rs-free method with a one-loop bispectrum up to k_Bmax = 0.23 h/Mpc. The paper's own Fig. 3 and Sec. III B show that adding the bispectrum changes the recovered h and Ωm and drives α_rs away from unity at 1.8σ, with sensitivity to the bispectrum scale cut. That is exactly the behavior one would expect if the broadband baryon suppression begins to carry measurable r_s information. I do not claim this proves the residual is present; it means the assumption is not yet validated for the new likelihood. The proposed r_s-injection test would settle it directly. The paper's other robustness checks—PyBird vs CLASS-PT agreement, prior-insensitivity, and the mock DDE/EDE tests—are genuine strengths, and no fraud or dishonesty is implied. The appropriate verdict remains CONDITIONAL, the same as the reader's, so no change is recommended.","tokens_in":19823,"tokens_out":10941,"duration_ms":124056,"concrete_test":"Perform an r_s-injection test on the EFTBOSS P+B likelihood at k_Bmax = 0.23 h/Mpc: generate the data vector from a fixed ΛCDM cosmology, then analyze it with the α_rs-free pipeline replacing only the no-wiggle template's baryon-suppression scale β r_s by ±2% and ±5% of its fiducial value, keeping all other inputs fixed. If the recovered h posterior shifts by more than ~0.01, or its width changes by more than ~10%, relative to the fiducial-template analysis, residual sound-horizon information in the broadband is non-negligible and the baseline h is not sound-horizon independent. A null shift would validate the Sec. II B assumption for the bispectrum-extended likelihood.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that α_rs marginalization in Eq. (1) removes sound-horizon dependence from the EFTBOSS likelihood, making the baseline h = 0.702 ± 0.022 a sound-horizon-free H0 measurement. The only direct support is the statement in Sec. II B that residual information through the baryon-suppression scale β r_s is negligible, citing Ref. [19]. But Ref. [19]'s demonstration was for power-spectrum-only analyses. Here the α_rs-free method is extended to the one-loop bispectrum with k_Bmax up to 0.23 h/Mpc (Sec. II A), and the paper's own Sec. III B shows this is exactly where the method becomes unstable: adding the bispectrum shifts h and Ωm by ~1σ, produces α_rs = 1.063 ± 0.035 (1.8σ from unity), and the shift largely disappears when k_Bmax is lowered to 0.18 h/Mpc. This is consistent with the small-scale bispectrum beginning to resolve broadband shape features—the channel through which r_s can re-enter P_nw after α_rs marginalization. If so, the baseline H0 error bars miss a systematic visible in the scale-cut dependence, and the 'sound-horizon-free' label is not established for this likelihood.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"Using the wiggle/no-wiggle decomposition of the linear matter power spectrum (Eq. 1) and marginalizing over the BAO rescaling parameter α_rs, the authors construct a 'sound-horizon-free' EFTofLSS likelihood from BOSS power spectrum and bispectrum at one loop, and combine it with DESI Legacy Survey DR9 angular C_ℓ's, DES Y3 3×2pt constraints (entering through Gaussian priors on {h, Ω_m, σ8}), and Planck PR3 CMB lensing, plus BBN and Planck priors on ω_b and n_s. The baseline gives h = 0.702^{+0.022}_{-0.024}, Ω_m = 0.310 ± 0.013, σ8 = 0.799 ± 0.020. Inclusion of Pantheon+ lowers h to 0.686 ± 0.018; adding a DESI DR2 BAO-based Ω_m prior yields h = 0.708^{+0.015}_{-0.017}. The EFTBOSS bispectrum analysis yields α_rs = 1.063 ± 0.035 at k_Bmax = 0.23 h/Mpc (1.8σ from unity), which becomes consistent with unity at k_Bmax = 0.18. The authors also run ΛCDM fits on EDE and DDE mocks and interpret the real-data α_rs-free/α_rs-fixed difference as consistent with evolving dark energy.","tokens_in":20199,"tokens_out":10619,"duration_ms":100179,"significance":"If valid, the result is one of the tightest sound-horizon-free H0 determinations from LSS alone (~3–4% precision), and the first to incorporate the BOSS bispectrum in the α_rs-marginalized framework. The paper's strengths include the multiprobe combination, the explicit internal-consistency quantification, the scale-cut study, the PyBird/CLASS-PT cross-check for the 2pt case, and the EDE/DDE mock tests. These are useful and go beyond many previous analyses. However, the sound-horizon-free label and the quoted precision are not yet supported by the analysis as presented: the bispectrum at the baseline scale cut introduces a scale-cut-dependent shift, and the baseline combination includes two likelihoods with ~2.6σ internal tension. These issues require revision before the central claims can be accepted.","major_comments":[{"comment":"The assertion that marginalizing α_rs fully removes sound-horizon information is borrowed from Ref. [19], which is a power-spectrum-only demonstration. The present paper extends this to the one-loop bispectrum with k_Bmax = 0.23 h/Mpc. Sec. III B and Fig. 3 show that this extension is not stable: at k_Bmax = 0.23, α_rs = 1.063 ± 0.035 (1.8σ from unity), and the α_rs-free vs α_rs=1 analyses differ by 1.8σ in h and 1.4σ in Ω_m; lowering k_Bmax to 0.18 reduces the discrepancy to 0.6σ and restores α_rs = 1. Since the small-scale bispectrum is exactly where broadband shape features (including the baryon-suppression scale β r_s) can re-enter P_nw after α_rs marginalization, the baseline EFTBOSS likelihood is not demonstrated to be sound-horizon-free. The quoted h error bars therefore miss a systematic that the scale-cut dependence reveals. Please make k_Bmax = 0.18 the baseline or propagate th","section":"Sec. II B, Eq. (1); Sec. III B, Fig. 3"},{"comment":"The baseline combination of EFTBOSS, DESICℓ and EFTDES contains internal tension: the paper reports a 2.6–2.7σ disagreement between EFTBOSS and EFTDES in the {h, Ω_m, σ8} space, using Eq. (2). These likelihoods are then multiplied as if statistically compatible. Combining two incompatible posteriors with different preferred regions (EFTBOSS h ≈ 0.63; EFTDES h ≈ 0.77) can shift the joint mode and artificially shrink uncertainties. The headline h = 0.702 ± 0.022 therefore depends on an unmodeled tension. The authors should either introduce a tension/shift parameter, report pairwise removals (e.g., baseline without EFTDES), or use a conservative hyperparameter combination. A robustness table showing that the central h and its error are stable under these choices is needed to support the claimed 3–4% precision.","section":"Sec. III A, Tab. I; Sec. III C"},{"comment":"The DES Y3 3×2pt contribution is not a full likelihood but Gaussian priors on {h, Ω_m, σ8} taken from Ref. [93]. This compression assumes a Gaussian posterior and discards correlations with other parameters such as ω_cdm and ln(10^10 A_s) that are free in the EFTBOSS analysis. Since EFTDES dominates the Ω_m constraint in the combination, a non-Gaussian or correlated tail could bias the final Ω_m and h. Please validate the Gaussian compression by comparing with the full likelihood in a common parameter space, or explicitly quantify the information loss; alternatively rerun the baseline with the full EFTDES likelihood.","section":"Sec. II A (EFTDES entry)"}],"minor_comments":[{"comment":"The redshift ranges for CMASS and LOWZ appear swapped: CMASS should be 0.43 < z < 0.7 (z_eff = 0.57) and LOWZ should be 0.2 < z < 0.43 (z_eff = 0.32). Please correct.","section":"Sec. II A"},{"comment":"The link for the public likelihood is incomplete ('here /github'). Provide the full URL.","section":"Sec. II A (DESICℓ)"},{"comment":"The y-axis labels for k_Bmax are easy to misread because the tick order is non-monotonic; consider reordering and adding explicit units.","section":"Sec. III B / Fig. 3"},{"comment":"The phrase '2.4% measurement on h free from sound horizon' should be 'free of'; also 'EFBOSS' appears as a typo in the bottom-left panel description.","section":"Sec. III C"},{"comment":"Define the abbreviations 'scf' and 'log10 ac' in the EDE row; currently only experts will recognize these parameters.","section":"Table II"},{"comment":"The DDE mock test shows a 1.9σ difference in one realization. The statement 'consistent with recent hints of DDE' is stronger than the evidence; please soften or provide multiple mock realizations / a p-value.","section":"Sec. III D"}],"recommendation":"major_revision","confidential_remarks":"The manuscript has a strong core and useful validation checks, but the central 'sound-horizon-independent' label is currently scale-cut dependent and the baseline combination includes unmodeled internal tension. The requested tests and reruns are within scope and should be feasible. I would be open to acceptance after major revision. The heavy self-citation of Ref. [84] for the DESICℓ likelihood and the incomplete public link should also be cleaned up."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this is the first sound-horizon-free H0 analysis to include the BOSS power spectrum and bispectrum, and the combination with DESI Cℓ, DES Y3, and Planck lensing gives a 3–4% h constraint that sits between Planck and SH0ES — a reasonable, useful result. Second, the paper is admirably transparent about its own soft spots: the bispectrum scale cut matters, and the authors show it explicitly.\n\nWhat is genuinely new: the joint multiprobe pipeline, the wiggle-no-wiggle procedure extended to the bispectrum, and the mock-based EDE/DDE diagnostic. The robustness checks are real: PyBird vs CLASS-PT in the alpha_rs-free context, prior sensitivity, and a tension metric between datasets. I believe the central h=0.702±0.023 is not a numerical artifact.\n\nThe main concern is the sound-horizon-free claim for the bispectrum. In Sec II B they dismiss residual r_s information through the baryon suppression scale by citing Ref [19], but that demonstration was for power spectra. Their own Fig 3 shows that adding bispectrum scales above k=0.18 shifts h and Ωm by ~1.4σ and makes alpha_rs deviate 1.8σ from unity; the deviation goes away at k=0.18. That is precisely the channel where r_s sensitivity could re-enter after alpha_rs marginalization. So I would not call the EFTBOSS bispectrum likelihood fully sound-horizon-free. The authors instead interpret this as a possible DDE signal, and their DDE mock does reproduce a similar shift — suggestive, but not yet a demonstration that the r_s channel is closed. This affects the label more than the headline number, since the full baseline returns alpha_rs=0.997±0.022, but the systematic sits in the core likelihood.\n\nOther soft spots, in proportion: the internal 2.6–2.7σ tension between EFTBOSS and EFTDES is reported but not modeled; the EFTDES contribution is a Gaussian prior approximation; and the DESI Cℓ likelihood comes from the authors' own preprint. None of these is disqualifying, and all are stated.\n\nBottom line: worth a serious referee. The paper is careful, reproducible in spirit (public likelihood), and the question it addresses matters. The referee should push on the bispectrum wiggle-no-wiggle validation, ideally with mocks that have known r_s-dependent features, and ask whether the combined posterior is robust under a tension-aware combination. I would accept after moderate revision.","headline":"A careful, transparent multiprobe analysis with a plausible headline H0, but the sound-horizon-free label for the BOSS bispectrum is not yet fully secured.","tokens_in":20755,"tokens_out":4048,"would_cite":true,"duration_ms":38763,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A sound-horizon-free multiprobe large-scale structure analysis measures H0 to 3–4% precision, yielding h = 0.702 ± 0.02.","keywords":["Hubble constant","sound horizon","large-scale structure","baryon acoustic oscillations","wiggle-no-wiggle split","H0 tension","dark energy","cosmological parameter estimation"],"falsifier":"A concrete test: vary the BBN prior on the baryon density omega_b (which sets the sound horizon) and see if the inferred H0 shifts by more than the quoted error. If shifting omega_b within its BBN uncertainty moves h by more than about 0.02, residual sound-horizon information is leaking through the broadband shape. Additionally, the alpha_rs deviation grows with the bispectrum scale cut; a measurement at higher k with larger survey volume would either confirm the deviation as real physics or reveal it as a modeling artifact.","tokens_in":1560,"feed_emoji":"🌌","tokens_out":3205,"duration_ms":86361,"temperature":0.7,"pith_summary":"This paper shows that the expansion rate of the universe can be estimated from the clustering of galaxies and other large-scale structure data without calibrating the cosmic sound horizon. It does this by splitting the matter power spectrum into a smooth part and a baryon acoustic oscillation part, then marginalizing over the BAO scale parameter. Combining the BOSS galaxy power spectrum and bispectrum with galaxy lensing, imaging clustering, and CMB lensing yields h = 0.702 with 3–4% precision. This matters because the Hubble tension depends on whether early-universe estimates share a common sound-horizon calibration, and the method provides an independent route. The paper also shows how the same analysis can diagnose new physics, finding consistency with hints of evolving dark energy.","feed_headline":"Sound-horizon-free galaxy maps measure H0 at 3–4% precision","feed_subtitle":"H0 = 0.702 from large-scale structure alone, avoiding the sound-horizon calibration at the heart of the Hubble tension.","key_machinery":"The wiggle-no-wiggle split: the linear matter power spectrum is decomposed into a smooth broadband component and an oscillatory BAO component, P_lin(k) = P_nw(k) + P_w(alpha_rs k). Marginalizing over alpha_rs removes the acoustic peaks that would otherwise calibrate the sound horizon. The method relies on the claim, from the literature, that sound-horizon information entering through the baryon suppression scale is negligible for current survey precision. This split is applied to the BOSS power spectrum and bispectrum, and the other probes are effectively sound-horizon-independent because their BAO signal is weak or projection-smeared.","core_discovery":"The paper claims that a sound-horizon-free multiprobe analysis of large-scale structure can determine H0, Omega_m, and sigma_8 at 3–4% precision. The key step is the wiggle-no-wiggle split of the linear matter power spectrum, P_lin(k) = P_nw(k) + P_w(alpha_rs k), with alpha_rs marginalized over so the BAO oscillation carries no weight. Combining the BOSS power spectrum and bispectrum (for the first time in this context) with the DESI angular power spectra, DES Y3 3x2pt, and Planck PR3 CMB lensing gives h = 0.702^{+0.022}_{-0.024}, Omega_m = 0.310 ± 0.013, sigma_8 = 0.799 ± 0.020. The paper also reports a 1.8σ deviation of alpha_rs from unity generated by bispectrum modes above k = 0.18 h/Mpc","pith_inferences":["If the wiggle-no-wiggle split fully removes sound-horizon information, the agreement between this H0 and Planck's sound-horizon-calibrated value is a nontrivial internal consistency check of Lambda CDM: the Hubble tension cannot be ascribed solely to sound-horizon systematics.","The sensitivity of the alpha_rs deviation to the bispectrum scale cut suggests that the effect could be a modeling artifact; future surveys with higher signal-to-noise will determine whether it is real physics or a consequence of pushing one-loop effective field theory too far.","The same principle could be used with the matter-radiation equality turnover as an independent standard ruler, providing an orthogonal check on the wiggle-no-wiggle results."],"forward_implications":["A sound-horizon-free H0 from LSS alone at ~3–4% precision, h = 0.702, lies between Planck and SH0ES, with 1.2σ and 2.1σ tensions respectively.","Adding Pantheon+ supernovae gives h = 0.686 ± 0.018 (2.6% precision), shifting the baseline toward lower H0 and higher Omega_m.","The BOSS bispectrum improves the h–Omega_m figure of merit by a factor of ~2 and induces a 1.8σ deviation of alpha_rs from unity; this deviation disappears when the bispectrum is restricted to k_B^max = 0.18 h/Mpc.","The sound-horizon-free analysis acts as a new-physics diagnostic: mock early-dark-energy cosmologies show no alpha_rs shift, while evolving-dark-energy mocks reproduce a ~1.9σ shift similar to the real data.","The methodology is directly applicable to upcoming surveys such as Euclid and LSST, which should yield tighter sound-horizon-independent constraints."],"fun_headline_variants":["H0 = 0.702 from sound-horizon-free large-scale structure","Sound-horizon-free H0: 3–4% precision from BOSS, DESI, DES Y3","Multiprobe LSS yields H0 without sound horizon","H0 from LSS alone: 0.702 ± 0.022, no sound horizon","Sound-horizon-free H0 measurement at 3–4% precision"],"cache_read_input_tokens":21888,"weakest_assumption_plain":"The analysis assumes that after marginalizing over alpha_rs, the remaining broadband shape of the power spectrum—in particular the baryon suppression scale proportional to the sound horizon—carries negligible sound-horizon information for current survey precision; if that residual is non-negligible, the 'sound-horizon-free' label on the H0 constraint fails.","fun_headline_variants_meta":{"raw":{"variants":["H0 = 0.702 from sound-horizon-free large-scale structure","Sound-horizon-free H0: 3–4% precision from BOSS, DESI, DES Y3","Multiprobe LSS yields H0 without sound horizon","H0 from LSS alone: 0.702 ± 0.022, no sound horizon","Sound-horizon-free H0 measurement at 3–4% precision"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000944,"raw_usage":{"total_tokens":3955,"prompt_tokens":913,"completion_tokens":3042,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":657,"completion_tokens_details":{"reasoning_tokens":2930}},"tokens_in":657,"tokens_out":3042,"duration_ms":19467,"temperature":1.0,"reasoning_tokens":2930,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T20:54:26.453209+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete test: vary the BBN prior on the baryon density omega_b (which sets the sound horizon) and see if the inferred H0 shifts by more than the quoted error. If shifting omega_b within its BBN uncertainty moves h by more than about 0.02, residual sound-horizon information is leaking through the broadband shape. Additionally, the alpha_rs deviation grows with the bispectrum scale cut; a measurement at higher k with larger survey volume would either confirm the deviation as real physics or reveal it as a modeling artifact.","supporting_citations":[],"review_version":1}