{"id":"0c494c86-543b-4362-bcec-8fd5b7334874","arxiv_id":"2604.04187","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":2,"one_line_summary":"XGBRegressor-optimized nonlocal TPA reportedly lowers RMS alpha half-life error 74.8% on 599 nuclei and predicts Z=117–120 half-lives in line with DZR and MUDL.","lead":"The abstract claims a machine-learning-tuned two-potential model cuts alpha-decay half-life errors by ~75% for hundreds of nuclei and then forecasts superheavy species. Only the abstract is available for this arXiv id; the attached full text is a different paper, so the numerical claims cannot be checked.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"Central 74.8% RMS claim cannot be audited: full text is a different paper, and the abstract gives no train/hold-out protocol for XGB-tuned nonlocal parameters.","rationale":"The reader correctly flagged abstract-only review, manuscript mismatch, and the risk that the 74.8% RMS drop is in-sample fitting of flexible nonlocal parameters rather than validated improvement. That is the single load-bearing soft spot for the strongest claim: without a documented hold-out or CV protocol, the metric does not support the physical conclusion. Multi-model agreement on Z=117–120 is secondary and does not rescue the fit quality claim. No independent tables or code are available here to overturn that concern, so the verdict stays UNVERDICTED with no upgrade. A concrete held-out recompute would settle whether the concern lands.","tokens_in":74219,"tokens_out":557,"duration_ms":9947,"concrete_test":"Obtain the nuclear manuscript (or code) for 2604.04187; recompute RMS on a held-out 20% of the 599 nuclei (or nested CV) with XGB hyperparameters fixed on the train fold only. If held-out RMS reduction falls well below ~75% (e.g. near original TPA), the headline improvement is in-sample fit, not out-of-sample physics.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing claim is that generalizing alpha-nucleus nonlocality to odd-A/odd-odd nuclei in TPA, with coordinate-dependent parameters fit by XGBRegressor, cuts RMS by 74.8% vs original TPA on 599 nuclei (Z=52–118), and that superheavy (Z=117–120) predictions then agree with DZR/MUDL. That improvement is only meaningful if the RMS is out-of-sample (or at least regularized and cross-validated). The abstract does not state train/validation/test splits, early stopping, feature set, or whether the 599 nuclei used to quote RMS are the same set on which XGB was fit. Flexible coordinate-dependent nonlocal parameters plus a gradient-boosting regressor can easily absorb residuals in-sample without improving the physical model. The supplied “full manuscript” is Moretti’s QFT localization paper (arXiv:2604.04173), not the nuclear TPA/XGB work, so equations, tables, residual plots, and protocol cannot be checked. Until splits and code exist, the 74.8% figure does not establish genuine physical improvement.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The submission is presented as a nuclear-theory paper that generalizes the alpha–nucleus nonlocality effect in the two-potential approach (TPA) to odd-A and odd-odd nuclei, optimizes coordinate-dependent nonlocality parameters with XGBRegressor, reports a 74.8% reduction in RMS half-life deviation relative to the original TPA on 599 nuclei (Z=52–118), and predicts half-lives for 142 superheavy nuclei (Z=117–120) in close agreement with the DZR and MUDL models. The supplied full manuscript text, however, is an unrelated mathematical-physics article (Moretti, arXiv:2604.04173) on relativistic spatial localization observables constructed from the stress–energy tensor in free QFT. None of the nuclear-physics claims, methods, or numerical results can be checked against the body of the document.","tokens_in":74537,"tokens_out":836,"duration_ms":11743,"significance":"If the abstract’s claims were supported by a matching, carefully validated manuscript, an improved TPA with a large RMS gain and consistent superheavy predictions would be of genuine interest for alpha-decay systematics and SHN half-life estimates. As submitted, that significance cannot be assessed: the body of the paper does not contain TPA, nonlocality parameters, XGBRegressor, the 599-nucleus sample, or the Z=117–120 predictions. The only load-bearing issue is the complete mismatch between title/abstract and full text.","major_comments":[{"comment":"Title, abstract, and arXiv identifier (2604.04187, nucl-th) describe a nuclear TPA/XGB study of alpha-decay half-lives, but the full manuscript is Moretti’s QFT localization paper (arXiv:2604.04173, math-ph). There is no TPA formalism, no nonlocality parameters, no XGBRegressor protocol, no 599-nucleus dataset, and no superheavy predictions. The central claims cannot be audited.","section":null},{"comment":"Even taking the abstract alone, the headline 74.8% RMS reduction is not accompanied by any train/validation/test split, cross-validation, regularization, or hold-out protocol for the XGB-optimized coordinate-dependent parameters. Without that protocol, the improvement may be an in-sample fit rather than a physical gain; the missing nuclear manuscript prevents any check of residual plots, feature set, or out-of-sample performance.","section":null},{"comment":"Agreement of improved TPA with DZR and MUDL for Z=117–120 is offered as support for the predictions, but no tables, uncertainties, or model definitions appear in the supplied text. Reliability of the 142-nucleus forecast cannot be evaluated.","section":null}],"minor_comments":[{"comment":"The abstract cites Medeiros [1], a previous study [2], DZR [3], and MUDL [4] without bibliographic detail in the provided materials; a correct nuclear manuscript would need a complete reference list and clear definitions of those models.","section":null},{"comment":"If a corrected nuclear manuscript is resubmitted, it should state explicitly whether the 599 nuclei used for the RMS figure are disjoint from the set used to train XGBRegressor, and should report hyperparameters and feature construction.","section":null}],"recommendation":"reject","confidential_remarks":"The cache/full-text pipeline has attached the wrong paper (Moretti 2604.04173) to abstract 2604.04187. This is not a borderline scientific disagreement; it is a document-identity failure. Reject is appropriate until the correct nuclear manuscript is provided. If the journal receives the real TPA/XGB paper, it should be treated as a new submission and reviewed on train/hold-out protocol and residual analysis, not on the present QFT text."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The one thing you need to know: I cannot review the actual manuscript of Hu & Wu (arXiv:2604.04187). What was supplied as full text is Moretti’s math-ph paper on relativistic localization in QFT (2604.04173). Everything below is abstract-only plus that mismatch.\n\nFrom the abstract alone, the contribution is incremental but real in its niche: they extend Medeiros-style alpha–nucleus nonlocality inside the two-potential approach to odd-A and odd-odd systems, fit the new coordinate-dependent nonlocality parameters with XGBRegressor, report a 74.8% RMS drop versus original TPA on 599 nuclei (Z=52–118), and then give half-life predictions for 142 superheavy candidates (Z=117–120) that sit close to DZR and MUDL (nearly identical to DZR). That is a legitimate applied-nuclear extension of an existing program, not a new framework.\n\nThe soft spot is load-bearing and not minor. Flexible, coordinate-dependent parameters plus a gradient-boosting regressor can absorb residuals on the same set used to quote RMS. The abstract never states train/validation/test splits, cross-validation, early stopping, feature list, or whether the 599-nucleus RMS is out-of-sample. Without that, the 74.8% figure does not establish physical improvement; it may be in-sample fit quality. Multi-model agreement on Z=117–120 is useful as a consistency check but does not fix the evaluation design.\n\nWho cares: people who already work alpha half-lives, TPA, and superheavy search planning. Nobody outside that strip needs this until the protocol is clear.\n\nI would not bring it to reading group until the real PDF (equations, residual plots, splits, code) is in hand. I would not cite it on present evidence. If a complete nuclear manuscript matching the abstract exists, a serious editor should still send it to referees—the program is standard enough and the claim is sharp enough to deserve a methods check—not desk-reject on abstract alone. Right now the central number is simply unauditable.","headline":"We only have the abstract for this nuclear TPA/XGB paper; the cached “full text” is Moretti’s unrelated QFT localization paper, so the 74.8% RMS claim cannot be audited.","tokens_in":75151,"tokens_out":559,"would_cite":false,"duration_ms":13991,"reading_group":"no","serious_thinker":"unclear","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"An improved two-potential model of alpha-nucleus nonlocality, tuned with gradient boosting, cuts half-life RMS error by 74.8% on 599 nuclei and predicts Z=117-120 superheavy decays in line with DZR and MUDL.","keywords":["alpha decay","half-lives","superheavy nuclei","two-potential approach","nonlocality","XGBRegressor","odd-A nuclei","odd-odd nuclei"],"falsifier":"A blinded or hold-out evaluation of the improved TPA on a reserved subset of the 599 known half-lives, or new measured half-lives for Z=117-120 nuclei that systematically deviate from the joint TPA-DZR-MUDL predictions.","tokens_in":75083,"feed_emoji":"⚛️","tokens_out":680,"duration_ms":6122,"temperature":0.7,"pith_summary":"This paper generalizes the alpha-nucleus nonlocality effect, previously treated mainly for even-even systems, to odd-A and odd-odd nuclei inside the two-potential approach. Coordinate-dependent parameters that encode that nonlocality are optimized with an XGBRegressor gradient-boosting model. On a set of 599 nuclei with Z from 52 to 118 the improved scheme lowers the root-mean-square deviation of calculated alpha-decay half-lives by 74.8 percent relative to the original TPA. The same calibrated model is then used, together with the independent DZR and MUDL formulas, to forecast half-lives of 142 superheavy candidates with Z=117-120; the three sets of predictions agree closely, and the improved TPA results are nearly identical to those of DZR. The practical claim is that a carefully tuned nonlocal correction can both describe known alpha emitters far more accurately and supply mutually consistent forecasts for the heaviest nuclei still under experimental study.","feed_headline":"Nonlocal TPA cuts alpha half-life error 74.8 percent","feed_subtitle":"Gradient-boosted nonlocality matches DZR for Z=117-120 superheavy forecasts","key_machinery":"The improved two-potential approach (TPA) that embeds a generalized, coordinate-dependent nonlocality effect for all parity classes and whose free parameters are fitted by XGBRegressor.","core_discovery":"Generalizing alpha-nucleus nonlocality to odd-A and odd-odd systems inside the two-potential approach, and optimizing the resulting coordinate-dependent parameters with gradient boosting, produces an improved TPA whose half-life RMS deviation on 599 nuclei (Z=52-118) is 74.8 percent lower than that of the original TPA, while its predictions for 142 superheavy nuclei (Z=117-120) stay in close agreement with the DZR and MUDL models.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["XGB-optimized nonlocal TPA slashes alpha half-life RMS by 74.8%","Nonlocality in TPA: 74.8% lower RMS on 599 nuclei half-lives","Gradient-boosted nonlocal TPA matches DZR for Z=117-120 alphas","Odd-A and odd-odd nonlocal TPA cuts half-life error 74.8%","Improved nonlocal TPA predicts superheavy alpha half-lives with DZR"],"cache_read_input_tokens":65664,"weakest_assumption_plain":"That the 74.8 percent RMS reduction reflects genuine physical improvement rather than an in-sample fit of flexible parameters on the same nuclei used to quote the metric, without a clear hold-out or validation protocol.","fun_headline_variants_meta":{"raw":{"variants":["XGB-optimized nonlocal TPA slashes alpha half-life RMS by 74.8%","Nonlocality in TPA: 74.8% lower RMS on 599 nuclei half-lives","Gradient-boosted nonlocal TPA matches DZR for Z=117-120 alphas","Odd-A and odd-odd nonlocal TPA cuts half-life error 74.8%","Improved nonlocal TPA predicts superheavy alpha half-lives with DZR"]},"model":"grok-4.5","effort":"low","cost_usd":0.002042,"raw_usage":{"total_tokens":881,"prompt_tokens":763,"num_sources_used":0,"completion_tokens":118,"cost_in_usd_ticks":20420000,"prompt_tokens_details":{"text_tokens":763,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":763,"tokens_out":118,"duration_ms":1332,"temperature":1.0,"reasoning_tokens":0,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-13T11:02:44.776002+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"A blinded or hold-out evaluation of the improved TPA on a reserved subset of the 599 known half-lives, or new measured half-lives for Z=117-120 nuclei that systematically deviate from the joint TPA-DZR-MUDL predictions.","supporting_citations":[],"review_version":1}