{"id":"c6602362-bcf9-4d12-a462-49a3a67f7595","arxiv_id":"2607.18797","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A charge-only maximum-likelihood reconstruction using calibration-derived nPE maps achieves ~4 cm vertex resolution and sub-1% relative energy-bias behavior for positrons in a ton-scale liquid scintillator detector.","lead":"A simulation study shows that a small liquid-scintillator neutrino detector can reconstruct an event's radius and energy from the pattern of charges seen by its silicon photomultipliers, using maps of expected light built from calibration sources. The method is a template-based likelihood fit that could be applied to any similarly sized spherical detector such as JUNO-TAO.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Claimed vertex performance (<2 cm bias, <4 cm resolution) is demonstrated only near the center; Fig. 6 shows bias ~30–40 mm and resolution ~50–60 mm at radii ≥650 mm, contradicting the unqualified conclusion.","rationale":"The central claim is the method's vertex performance. The reader's weakest assumption focused on template mismatch between simulation and real calibration. My concern is more direct: even within the same simulation, the published figures contradict the unqualified conclusion. This is a concrete, internal inconsistency that can be settled by re-computing radial/volume-averaged metrics. It does not overturn the paper's contribution—the method still works well near the center—but it requires the authors to qualify the claimed performance and report where it holds. The reader's CONDITIONAL verdict remains appropriate; this concern adds a specific condition about radial coverage rather than changing the overall verdict.","tokens_in":12073,"tokens_out":7891,"duration_ms":80478,"concrete_test":"Use the existing MC samples to compute, for each positron kinetic energy, the radius r* at which the radial resolution first exceeds 40 mm; if r* < 650 mm, the unqualified conclusion fails within the intended fiducial volume. Also compute the volume-weighted average resolution ∫ r² σ_R(r) dr / ∫ r² dr over r ∈ [0,650] mm and the fraction of events with bias <2 cm and resolution <4 cm, and report these alongside the center values.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The Conclusion states 'vertex bias less than 2 cm and vertex resolution less than 4cm can be achieved for the positrons' without radial qualification. However, Fig. 6 shows this holds only near the center of the detector. For positrons, the radial resolution (RMS of R_rec − R_QEdep) exceeds 40 mm for radii ≳500–600 mm at all kinetic energies, and the bias approaches ~30–40 mm near the 650 mm boundary of the nominal fiducial volume (Sec. IV B refers to 'edge of the fiducial volume' at 650 mm). Table II gives center values of 15–16 mm bias and 38–38.5 mm resolution—the resolution only 2 mm below the claimed 4 cm. Because IBD vertices are roughly uniform in the liquid-scintillator volume, the volume-weighted resolution will be substantially worse than 4 cm. This is an internal inconsistency in the reported performance, not a matter of external calibration. The paper should report performance as a function of radius and specify the fiducial radius over which the claim holds.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a charge-only maximum-likelihood reconstruction method (QMLE) for simultaneous reconstruction of the radius and visible energy of IBD positrons in a ton-scale liquid scintillator detector of the JUNO-TAO type. The method builds templates of the expected number of photoelectrons per MeV as a function of event radius and SiPM angle, using simulated calibration sources (68Ge, 137Cs), interpolates these nPE maps between calibration radii and along axes, and then minimizes a Poisson-Gaussian charge likelihood over r and E, with θ and φ fixed from the CCA algorithm. The authors report Monte Carlo results for reconstructed radial bias and resolution for e+ and e− samples and study the effect of reflections, cross-talk, dark noise, and quenching on energy resolution. The conclusion states that vertex bias below 2 cm and vertex resolution below 4 cm can be achieved for positrons, and that the method is applicable to similar spherical neutrino detectors.","tokens_in":12279,"tokens_out":4114,"duration_ms":44628,"significance":"If the claimed performance were fully substantiated, the method would be a useful contribution to reconstruction in compact liquid scintillator detectors: it uses only charge information, exploits spherical symmetry, and is intended to be calibrated from radioactive source data rather than from an analytical optical model. The template-based charge likelihood is a reasonable and transparent statistical framework, and the paper includes MC samples covering multiple energies and radii, and some robustness checks on cross-talk, dark noise, and reflections. However, the paper as presented does not support its headline claims: no energy resolution number is reported despite the abstract promising 'unprecedented energy resolution', and the vertex performance is qualified only in a small central region, not over the fiducial volume as the conclusion implies. These issues are central and require additional analysis or a substantial reframing of the claims.","major_comments":[{"comment":"The conclusion states 'vertex bias less than 2 cm and vertex resolution less than 4 cm can be achieved for the positrons' without radial qualification. Fig. 6(b) shows this holds only for small radii; for radii ≳500 mm the radial resolution is above 40 mm at all kinetic energies, and the bias grows to ~30–40 mm near the 650 mm fiducial boundary. Table II gives center values of 15–16 mm bias and 38–38.5 mm resolution for e+, so even at the center the resolution is only 2 mm below the claimed 4 cm. Because IBD vertices are roughly uniformly distributed in the LS volume, the volume-averaged resolution will be substantially worse than 4 cm. The paper should report the claimed performance as a function of radius and specify the fiducial radius over which the 'less than 4 cm' statement is valid.","section":"Sec. V, Sec. IV A, Fig. 6, Table II"},{"comment":"The abstract's central promise is 'unprecedented energy resolution', but the paper never quotes an energy resolution value. Sec. IV B reports only relative changes ('less than 1%', 'around 0.5%', 'around 0.16%') and energy bias in Figs. 9 and 10; there is no σ(E_rec)/E or equivalent resolution metric anywhere. Without this number, the energy-reconstruction claim is unsupported. The authors should add energy resolution as a function of radius and energy for both nPE map types, and if the energy resolution is not competitive with existing methods, the abstract should be revised accordingly.","section":"Abstract, Sec. IV B, Figs. 9–10"},{"comment":"The performance numbers are obtained in a fully self-consistent MC setup: both the nPE-map templates and the reconstructed event samples are generated with the same Geant4 simulation and the same optical model. This is a closure test, not a validation against real calibration data. The paper emphasizes that templates will be built from calibration sources, but it does not test sensitivity to template mismatch—e.g., interpolation errors between the five calibration radii, imperfect source-position knowledge, or differences between simulated and actual reflection/absorption. A concrete test would be to perturb the optical parameters or source positions when building templates and then reconstruct the nominal samples, showing that the quoted biases and resolutions are stable at the claimed level.","section":"Sec. III, Figs. 3–4, Sec. IV"}],"minor_comments":[{"comment":"The manuscript contains many typos and grammatical errors, e.g., 'scintilling', 'the the', 'achi-eved', 'reletive', 'infromation', 'reconstrucede +', 'met-hod', and 'eutrinos'. A thorough language edit is needed.","section":"Throughout"},{"comment":"The y-axis label in Figs. 9 and 10 reads 'Mean(E_vis − E_rec_vis)' but the panel title says 'QMLE Vertex Bias'. This is misleading; it should be labeled 'energy bias'.","section":"Figs. 9–10"},{"comment":"The relative changes 'less than 1%', 'around 0.5%', and 'around 0.16%' are quoted without statistical uncertainties or the baseline resolution values to which they refer. Please provide the absolute resolutions and errors for these comparisons.","section":"Sec. IV B"},{"comment":"The calibration system description would benefit from a clear statement of the radius coverage of the ACU and CLS calibration points, and how the five radii in Fig. 3 (0, 200, 375, 650, 850 mm) correspond to the actual calibration path.","section":"Sec. II / Fig. 1"},{"comment":"The notation 'unhit' and 'hit' is not formally defined; please state that unhit channels are those with zero recorded photoelectrons. Also, the Gaussian charge response in Eq. (2) uses the same S_i for all k; its validity for small k should be commented on.","section":"Eq. (4)"},{"comment":"The interpolation from five calibration radii to arbitrary radii is stated to work 'similar' for cubic and polynomial methods, but no quantitative comparison is shown. It would help to report the interpolation residuals or to state that the results are insensitive to the interpolation order.","section":"Sec. III"}],"recommendation":"major_revision","confidential_remarks":"The paper reads like an early draft and the central claims are not backed by the reported data. The missing energy resolution number and the unqualified vertex-resolution conclusion are the two main blockers. The method itself is plausible, and the MC-based framework could become publishable if the authors add the missing performance metrics, qualify the radial dependence, and include a template-mismatch robustness test. I would not accept the manuscript in its current form, but these issues are fixable within the scope of the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nYou should know before reading: this is a capable, honest adaptation of Wu et al.'s charge-likelihood reconstruction (Ref [26]) to a small spherical LS detector like JUNO-TAO. The new piece is radius reconstruction as a second template parameter, with nPE maps built from simulated 137Cs and 68Ge calibration sources. That is genuinely useful for a ton-scale detector where full 3D vertex reconstruction may be unnecessary.\n\nWhat it does well: the MC setup is serious (Geant4, full optical model, SiPM layout, missing tiles, crosstalk and dark noise). The comparison of 137Cs vs 68Ge templates and the energy-dependent bias studies are honest. The robustness checks on reflections, crosstalk, and dark noise, while only quoted as relative changes, show the authors are thinking about the right systematics. The paper also credits Ref [26] clearly rather than overclaiming novelty.\n\nThe soft spots are real but fixable. The abstract promises 'unprecedented energy resolution,' yet the energy reconstruction section only shows bias plots—no absolute energy resolution number anywhere. That is load-bearing; for this detector class, energy resolution is the point. The robustness checks ('less than 1%', 'around 0.5%', 'around 0.16%') are meaningless without the baseline.\n\nThe vertex conclusion is overstated. 'Vertex bias less than 2 cm and vertex resolution less than 4 cm for positrons' holds only near the center. In Fig. 6, the resolution crosses 40 mm around 400–500 mm radius and the bias grows to 30–40 mm near the 650 mm fiducial boundary. Since IBD vertices are roughly volume-uniform, the volume-weighted resolution will be significantly worse than 4 cm. The conclusion needs a radius range attached.\n\nOne more limitation: the method reconstructs radius and energy while taking theta and phi from the CCA. That's a reasonable division of labor, but it is not a full 3D vertex reconstruction, and the title shouldn't imply otherwise.\n\nFinally, the templates and event samples come from the same MC, so the quoted performance is a self-consistency measure. A mismatched-template test (e.g., changed optical parameters) would tell us much more about robustness.\n\nBottom line: this is a useful incremental paper for the TAO/ton-scale LS community. It deserves a serious referee, but it needs an absolute energy-resolution table, radial-binned vertex performance, and a mismatched-template test before I'd accept it. I'd rather cite Wu et al. for the method, but I'd bring this to a reading group as a good example of a template-based likelihood applied to a small detector.\n\nRecommendation: send to peer review, not desk reject, with the expectation that the missing quantification is required for acceptance.","headline":"Honest but overstated application of a known likelihood template method to a small LS detector; the missing energy-resolution number and radius-dependent vertex claim need fixing before the headline results can be trusted.","tokens_in":12825,"tokens_out":4210,"would_cite":false,"duration_ms":40332,"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 charge-only maximum-likelihood method using calibration-derived nPE maps reconstructs both radius and visible energy of positron events in a ton-scale spherical liquid-scintillator detector, achieving in Monte Carlo a vertex bias below 2","keywords":["liquid scintillator detector","SiPM","charge reconstruction","nPE map","maximum likelihood","vertex reconstruction","energy reconstruction","reactor antineutrino"],"falsifier":"Take a real calibration source at a radius not in the template grid, say 500 mm, and reconstruct it using templates built from the other radii; if the reconstructed radius bias or resolution exceeds the quoted few-centimeter values, the linear-interpolation or spherical-symmetry assumption breaks. Separately, compare real 137Cs/68Ge calibration maps with simulated maps: a mismatch in the reflection peak or angular shape would directly invalidate the MC-based performance numbers.","tokens_in":11883,"feed_emoji":"⚛️","tokens_out":6096,"duration_ms":53650,"temperature":0.7,"pith_summary":"The paper tries to establish that a compact spherical liquid-scintillator detector can reconstruct an event's radius and visible energy using only the charge collected by SiPMs, without timing information. The method builds expected 'number of photoelectrons per MeV' maps from simulated calibration sources, interpolates them to arbitrary radii, and maximizes a Poisson-Gaussian likelihood over radius and energy. In Monte Carlo, positron vertices are reconstructed with bias below 2 cm and resolution below 4 cm, with energy bias growing mainly from energy leakage near the detector edge. If the method holds on real data, it would give near-detector reactor experiments a simple, calibration-driven way to measure antineutrino spectra precisely while keeping instrumentation minimal.","feed_headline":"Charge-only likelihood reconstructs radius and energy to within 4 cm","feed_subtitle":"Calibration-built nPE maps let a compact detector extract radius and energy from charge alone.","key_machinery":"The central object is the nPE map, mu_hat(r, theta_SiPM): the expected number of photoelectrons per MeV of visible energy seen by each SiPM as a function of the event radius r and the angle between the event position and the SiPM direction. It feeds a likelihood that combines a Poisson distribution for the number of photoelectrons on each channel with a Gaussian single-photoelectron charge response; maximizing the negative log-likelihood over r and E gives the reconstructed values. The map's separability into radius and angle plus linear interpolation between calibration points is what carries the argument.","core_discovery":"On its own terms, the paper claims that the expected light pattern on the SiPM array, expressed as an nPE map per unit visible energy and parameterized only by event radius and SiPM polar angle, is a sufficient statistic for simultaneous radius and energy reconstruction. Because the detector is spherical, maps measured along one axis describe all directions, and maps at five calibration radii can be linearly interpolated. The maximum-likelihood fit over radius and energy, with angles fixed by a center-of-charge seed, yields sub-4-cm radius resolution and sub-2-cm bias for positrons in simulation, and an energy resolution whose dominant term is photoelectron statistics, with quenching the lar","pith_inferences":["A natural test beyond the paper is to build nPE maps from real calibration data at the five radii and reconstruct events from an independent source position; success would confirm that linear interpolation and one-axis maps hold in the physical detector.","Because the quoted performance comes from the same simulated detector used to produce the templates, a mismatch between simulation and the real optical response (e.g., unmodeled reflections or SiPM non-uniformity) would change the numbers; the method's robustness to such mismatch is not yet demonstrated.","The radial resolution for positrons is worse than for electrons because annihilation gammas smear the light profile; combining charge-only reconstruction with timing information could push vertex resolution below the quoted 4 cm.","The energy bias rising with energy near the boundary points to energy leakage; an energy-dependent template interpolation or an explicit leakage correction could extend the method's useful fiducial volume."],"forward_implications":["Charge-only information is sufficient for radial and visible-energy reconstruction in compact spherical liquid-scintillator detectors, so timing readout is not needed for these quantities.","Calibration with gamma sources at a few positions (e.g., 137Cs and 68Ge) can generate templates that work across event energies; the energy-dependent reflection region changes the overall resolution by less than 1%.","The same template method transfers to other detectors of similar size and spherical geometry, provided their SiPM response can be calibrated.","The main limit to energy resolution is photoelectron statistics; crosstalk, dark noise, and reflections each contribute roughly half a percent or less, while scintillator quenching is the largest non-statistical effect."],"fun_headline_variants":["Charge-only likelihood locates events to 4 cm in compact detector","SiPM charge fit yields sub-4-cm radius in ton-scale scintillator","Radius and energy from charge maps alone, no direction data","Tiny detector, tight radii: charge-only likelihood achieves 4 cm","Charge-based template method reaches 4-cm precision in LS detector"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The method assumes that nPE maps measured at five calibration radii can be linearly interpolated to any radius, and that maps measured along one axis describe all directions because the detector is perfectly spherically symmetric.","fun_headline_variants_meta":{"raw":{"variants":["Charge-only likelihood locates events to 4 cm in compact detector","SiPM charge fit yields sub-4-cm radius in ton-scale scintillator","Radius and energy from charge maps alone, no direction data","Tiny detector, tight radii: charge-only likelihood achieves 4 cm","Charge-based template method reaches 4-cm precision in LS detector"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000285,"raw_usage":{"total_tokens":1527,"prompt_tokens":769,"completion_tokens":758,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":513,"completion_tokens_details":{"reasoning_tokens":679}},"tokens_in":513,"tokens_out":758,"duration_ms":8314,"temperature":1.0,"reasoning_tokens":679,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T14:15:37.897994+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a real calibration source at a radius not in the template grid, say 500 mm, and reconstruct it using templates built from the other radii; if the reconstructed radius bias or resolution exceeds the quoted few-centimeter values, the linear-interpolation or spherical-symmetry assumption breaks. Separately, compare real 137Cs/68Ge calibration maps with simulated maps: a mismatch in the reflection peak or angular shape would directly invalidate the MC-based performance numbers.","supporting_citations":[],"review_version":1}