{"id":"6908b99a-3db5-458e-88fd-fce3a01237e2","arxiv_id":"2511.03910","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Neural posterior estimation reconstructs neutrino energy, direction, and flavor from radio signals in ice detectors, reporting median resolutions of 0.30 log(E) and 18 sq deg (shallow) and 0.08 log(E) and 28 sq deg (deep) for NC events at 1 EeV.","lead":"The paper presents a deep neural network using conditional normalizing flows to reconstruct neutrino direction, energy of the induced shower, and event topology from raw radio waveforms in in-ice detectors. This enables full posterior distributions for uncertainties and reported improvements in resolution for ultra-high-energy neutrino events.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Simulation fidelity for training data is the load-bearing assumption; GoF metric does not yet demonstrate that NPE posteriors remain calibrated under realistic unmodeled effects","rationale":"The reader's weakest assumption matches the central risk exactly. The paper's inclusion of a GoF score and some systematic studies is positive but does not close the loop on whether the NPE outputs remain reliable outside the training manifold; the concrete test above would directly probe that gap without requiring real data.","tokens_in":1799,"tokens_out":367,"duration_ms":25783,"concrete_test":"Generate a test set of 5000 events at 1 EeV with one additional systematic (e.g., ice attenuation length varied by ±15 % outside the range used in training) and recompute the median angular and energy resolutions plus posterior coverage; if either resolution degrades by >25 % or coverage deviates from nominal by >10 % at 68 % credible intervals, the generalization claim is not supported.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The headline performance numbers (0.30 log(E), 18 deg² shallow; 0.08 log(E), 28 deg² deep for 1 EeV NC) and the claim of event-by-event posterior PDFs via conditional normalizing flows are obtained exclusively on Monte Carlo. The paper quantifies selected systematics and introduces a goodness-of-fit score for compatibility with training simulations, yet this does not directly test whether the learned posterior remains accurate when detector response, ice properties, or noise realizations deviate from the simulated distribution in ways not spanned by the training variations. Because NPE performance is a direct function of how well the conditional density matches the true data-generating process, any unaccounted domain shift would invalidate both the quoted resolutions and the uncertainty estimates.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript presents a deep neural network using conditional normalizing flows for neural posterior estimation (NPE) to reconstruct neutrino direction, energy, and event topology (including flavor via CC/NC classification) from raw radio waveforms in in-ice detectors. It reports improved median resolutions over prior algorithms—0.30 log(E) and 18 deg² (shallow) and 0.08 log(E) and 28 deg² (deep) for 1 EeV NC events—along with full posterior PDFs for event-by-event uncertainties, quantification of antenna and systematic effects, and a goodness-of-fit score for MC compatibility.","tokens_in":1976,"tokens_out":547,"duration_ms":31575,"significance":"If the MC-derived posteriors and resolutions generalize, the work would advance reconstruction for UHE radio neutrino arrays by enabling calibrated uncertainty estimates and handling of stochastic ν_e CC events, strengthening sensitivity projections for experiments at the South Pole and Greenland. The explicit use of normalizing flows for full posteriors and the GoF metric are positive steps toward reproducible, uncertainty-aware analysis.","major_comments":[{"comment":"Abstract and results section: the headline resolutions (0.30 log(E), 18 deg² shallow; 0.08 log(E), 28 deg² deep for 1 EeV NC) and the claim of calibrated event-by-event posteriors are obtained exclusively on the Monte Carlo training distribution; the introduced goodness-of-fit score tests compatibility with the same simulations but does not directly probe calibration under unmodeled domain shifts in detector response, ice properties, or noise realizations outside the spanned variations.","section":"Abstract and results section"},{"comment":"Validation and systematics discussion: while selected systematics are quantified, no independent hold-out dataset, real-data proxy, or stress test of posterior coverage under realistic mismatches is presented, leaving the load-bearing assumption that the learned conditional density matches the true data-generating process unverified for deployment on measured signals.","section":"Validation and systematics discussion"}],"minor_comments":[{"comment":"Define 'shallow' and 'deep' detector components explicitly in the introduction or methods before quoting component-specific resolutions.","section":"Introduction"},{"comment":"Add explicit details on training/validation splits, hyperparameter selection, and how uncertainties are propagated from the flow to the reported median resolutions.","section":"Methods"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a reasonable fit for astro-ph.IM; the simulation-only validation is the primary concern but appears addressable with additional tests within the current scope."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for their constructive and detailed report. We address each major comment below, indicating where revisions have been made to the manuscript.","responses":[{"response":"We agree that the headline resolutions and posterior calibration results are evaluated on the Monte Carlo training distribution. This is a standard limitation in the field, as no real UHE neutrino events have been recorded by these detectors to date. The goodness-of-fit score is explicitly designed to test per-event compatibility with the training simulations and can flag potential mismatches. In the revised manuscript we have expanded the discussion of these limitations, including the assumptions about domain shifts and the role of the GoF metric in future analyses with real data.","revision_made":"partial","referee_comment":"[Abstract and results section] Abstract and results section: the headline resolutions (0.30 log(E), 18 deg² shallow; 0.08 log(E), 28 deg² deep for 1 EeV NC) and the claim of calibrated event-by-event posteriors are obtained exclusively on the Monte Carlo training distribution; the introduced goodness-of-fit score tests compatibility with the same simulations but does not directly probe calibration under unmodeled domain shifts in detector response, ice properties, or noise realizations outside the spanned variations."},{"response":"We acknowledge that the manuscript does not include an independent real-data hold-out set or explicit stress tests for unmodeled mismatches outside the simulated variations. Selected systematics (antenna types, ice properties) are quantified within the ranges covered by the training simulations. In the revision we will add controlled stress tests that perturb noise realizations and ice parameters beyond the training distribution to evaluate posterior coverage under such mismatches, thereby providing a clearer assessment of robustness.","revision_made":"yes","referee_comment":"[Validation and systematics discussion] Validation and systematics discussion: while selected systematics are quantified, no independent hold-out dataset, real-data proxy, or stress test of posterior coverage under realistic mismatches is presented, leaving the load-bearing assumption that the learned conditional density matches the true data-generating process unverified for deployment on measured signals."}],"tokens_in":1470,"tokens_out":485,"duration_ms":35265,"standing_objections":["Direct validation of posterior calibration and resolution on actual measured UHE neutrino signals, as no such events have been observed with in-ice radio detectors to date."]},"desk_editor":{"model":"grok-4.3","letter":"The main point is that this paper applies conditional normalizing flows to produce full posterior PDFs for neutrino energy and direction from raw radio waveforms in ice. That gives per-event uncertainties rather than single-point guesses, and they report median resolutions of 0.30 log(E) and 18 square degrees for a shallow component and 0.08 log(E) and 28 square degrees for a deep one on 1 EeV neutral-current events. They also reconstruct the more variable electron-neutrino charged-current cases and test antenna variations plus some systematics. A goodness-of-fit score is added to flag when a measured signal looks incompatible with the training simulations. These pieces together are new for this detector class. The work improves on earlier regression-style methods and supplies a concrete tool that future arrays could run on raw data. The numbers are presented clearly and the flow architecture is a reasonable choice for density estimation here. The GoF metric is a practical safeguard that earlier papers often skipped. All of this is done on Monte Carlo, with the training and test sets drawn from the same simulation family. The central resolutions and uncertainty calibration therefore depend on how faithfully those simulations capture real ice attenuation, noise statistics, and interaction details. The GoF can catch gross mismatches but does not automatically guarantee that the learned posteriors remain accurate under domain shifts not spanned by the training variations. That assumption is load-bearing for any claim about real-event performance. This paper is for groups already running or planning radio arrays at the South Pole or in Greenland who need reconstruction code that returns uncertainties. Readers focused on machine-learning methods for sparse, high-energy signals will see the most direct value. It deserves a serious referee because the method is grounded, the application is timely, and the GoF addition is a clear step forward even if further validation on independent data would be needed before deployment.","headline":"Conditional normalizing flows deliver full posteriors for radio neutrino direction and energy on simulations, with a useful GoF check, but the quoted resolutions rest on untested sim-to-data fidelity.","tokens_in":2462,"tokens_out":447,"would_cite":false,"duration_ms":21030,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"A neural network with conditional normalizing flows reconstructs neutrino energy, direction and flavor from radio waveforms, predicting full posterior distributions for uncertainties.","keywords":["neutrino reconstruction","radio detection","in-ice detectors","neural posterior estimation","normalizing flows","ultra-high-energy neutrinos","event topology","posterior PDF"],"falsifier":"A large, statistically significant discrepancy between the posterior distributions predicted by the network on real data and the distributions expected from independent Monte Carlo simulations of the same detector configuration.","tokens_in":2702,"feed_emoji":"📡","tokens_out":768,"duration_ms":26055,"temperature":0.7,"pith_summary":"The paper develops a deep neural network to extract neutrino direction, shower energy and interaction topology directly from raw radio antenna waveforms in in-ice detectors. It applies neural posterior estimation with conditional normalizing flows to output the complete probability distribution over energy and direction for each event, rather than single point estimates. This yields improved median resolutions of 0.30 in log(E) and 18 square degrees for shallow components and 0.08 in log(E) and 28 square degrees for deep components on neutral-current events at 1 EeV. The method also reconstructs the more variable charged-current electron-neutrino events and supplies a goodness-of-fit score to check whether measured signals are consistent with the Monte Carlo training set. A sympathetic reader would care because event-by-event uncertainties and better handling of stochastic topologies could tighten limits on ultra-high-energy neutrino fluxes and sources.","feed_headline":"Neural flows output full posteriors for neutrino energy and direction","feed_subtitle":"Conditional normalizing flows deliver event-by-event uncertainties and improved resolutions for both shallow and deep radio detector arrays.","key_machinery":"Conditional normalizing flows inside a neural posterior estimation network that model the full posterior distribution of neutrino parameters conditioned on the recorded radio waveforms.","core_discovery":"A deep neural network trained on Monte Carlo simulations reconstructs the neutrino direction, the energy of the induced particle shower, and the event topology from raw radio waveforms. For the first time the network outputs the full posterior probability density function for energy and direction by means of conditional normalizing flows, which directly supplies per-event uncertainty estimates. On neutral-current events at a shower energy of 1 EeV the approach achieves a median resolution of 0.30 in log(E) and 18 square degrees for a shallow detector component and 0.08 in log(E) and 28 square degrees for a deep component, outperforming earlier reconstruction algorithms while also handling st","pith_inferences":["The same architecture could be retrained on hybrid optical-radio data sets to cross-calibrate energy scales between detector technologies.","If the posterior widths prove reliable, they could be used to weight events in source-association studies without additional simulation campaigns.","Detector design studies could replace slow template-fitting reconstructions with this fast network to scan larger parameter spaces of antenna spacing and depth."],"forward_implications":["Event-by-event uncertainty estimates become available for downstream statistical analyses of ultra-high-energy neutrino fluxes.","Reconstruction extends to the more stochastic charged-current electron-neutrino events that were previously difficult to handle.","The impact of different antenna types and systematic uncertainties on resolution can be quantified directly from the network output.","A goodness-of-fit score derived from the posterior allows rejection of events whose waveforms are incompatible with the training simulations."],"fun_headline_variants":["Neural flows give neutrino energy and direction posteriors","Conditional flows enable per-event uncertainty for neutrino reconstruction","Deep nets improve resolutions for in-ice radio neutrino detection","Posterior PDFs for neutrino properties from raw radio antenna signals"],"cache_read_input_tokens":64,"weakest_assumption_plain":"The Monte Carlo simulations used for training accurately capture all relevant detector responses, neutrino interaction physics, and systematic effects so that the trained model generalizes to real measured signals.","fun_headline_variants_meta":{"raw":{"variants":["Neural flows give neutrino energy and direction posteriors","Conditional flows enable per-event uncertainty for neutrino reconstruction","Deep nets improve resolutions for in-ice radio neutrino detection","Posterior PDFs for neutrino properties from raw radio antenna signals"]},"model":"grok-4.3","cost_usd":0.008956,"raw_usage":{"total_tokens":4058,"prompt_tokens":736,"num_sources_used":0,"completion_tokens":60,"cost_in_usd_ticks":89562000,"prompt_tokens_details":{"text_tokens":736,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3262,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":736,"tokens_out":60,"duration_ms":66003,"temperature":1.0,"reasoning_tokens":3262,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-05-18T00:29:14.225782+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A large, statistically significant discrepancy between the posterior distributions predicted by the network on real data and the distributions expected from independent Monte Carlo simulations of the same detector configuration.","supporting_citations":[],"review_version":1}