{"id":"3d3ed71d-a1d8-4da0-8bfa-f2b6ece61189","arxiv_id":"2508.08052","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"The abstract and full text are two different papers: the abstract describes continual learning (CLEMC), while the body covers AugerPrime cosmic-ray detectors, so the claimed results are unverifiable from this submission.","lead":"This submission's abstract describes a theoretical model of model capacity in continual learning, but the full text is an unrelated paper about the AugerPrime cosmic-ray observatory. Because the body does not match the abstract, the continual-learning claims cannot be evaluated from the provided text.","discovery_kind":"unclear","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The submitted full text (an AugerPrime/ICRC2025 astroparticle paper) contains none of the CLEMC continual-learning theory, the promised difference equation, or any experiments on NN capacity; the abstract's central claim cannot be checked or validated from the provided submission.","rationale":"The reader's verdict of UNVERDICTED is correct because the submitted text does not contain the substance of the claimed contribution. My stress-test confirms this and sharpens it: the full text is entire unrelated to the abstract—it is a Pierre Auger Collaboration paper on AugerPrime, not a continual-learning paper. The strongest claim in the abstract is a broad generalization about neural network capacity under distribution shift. That claim depends critically on the difference-equation model being a faithful representation of learning dynamics. Without the equation's definition, derivation, assumptions, and validation, we cannot assess whether the non-stationarity result is a real phenomenon or an artifact of the model's construction. The manuscript as received cannot support the central claim. This is not an ad hominem concern; it is a structural failure of the submission. No adjustment to the reader's verdict is needed—UNVERDICTED is the appropriate status. I agree with the reader's weakest_assumption, and the concrete test above would settle whether the concern is a mere artifact of document extraction or a genuine absence of the paper's core content.","tokens_in":2338,"tokens_out":2539,"duration_ms":27980,"concrete_test":"Perform a software search of the full text for the tokens 'CLEMC', 'continual', 'capacity', 'stability-plasticity', 'difference equation', 'neural network', 'transformer', and 'GNN'. Also check the arXiv submission for supplementary files or additional pages beyond the acknowledgment section. If none of these tokens appear and no supplementary material exists, the submission is a mismatched/empty paper, and the verdict remains UNVERDICTED. If the tokens do appear, read the relevant sections to verify whether the CLEMC derivation, the difference equation, and the experimental results are actually present; if they are, the reader's complaint must be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract's strongest claim is a universal, architecture-independent law: a NN's ability to represent new tasks diminishes whenever incoming task distributions differ from previous ones. To support this, the paper must (1) define CLEMC, (2) derive the difference equation governing the interplay between NN, data, and optimization, (3) state its assumptions and regime of validity, and (4) validate it experimentally across the listed architectures (MLPs, CNNs, GNNs, transformer-based LLMs). The provided full text is 'AugerPrime: Status and first results'—a detector paper for ultra-high-energy cosmic rays. It contains no mention of continual learning, neural networks, capacity, stability-plasticity, CLEMC, or any related equations or experiments. This is a complete mismatch between abstract and body. Consequently, the central theoretical derivation and empirical support are entirely absent. The universal claim cannot be checked: we cannot determine whether it follows from the model, whether the model is reasonable, or whether any experiments support it. This is not a subtle under-specification; it is a different paper. The burden is on the authors to supply the actual content. Without it, the central claim is unsupported by the submitted full text.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript, as submitted under the stated arXiv identifier, consists of an abstract announcing a theory of continual learning (introducing 'CLEMC', a difference equation for capacity dynamics, and a claim that effective capacity is non-stationary in an architecture-independent way) followed by a full text that is an AugerPrime astroparticle detector status paper (ICRC2025) with no connection to continual learning, neural networks, capacity, or the abstract's content. The body contains no definitions, derivations, equations, experiments, or references relevant to the claimed contribution.","tokens_in":2653,"tokens_out":1132,"duration_ms":14563,"significance":"If the claims in the abstract were substantiated, the paper would offer a general, architecture-independent law for stability-plasticity dynamics in continual learning, which would be a notable theoretical contribution. The abstract promises a formal quantity (CLEMC), a difference equation, theoretical guarantees, and experiments across MLPs, CNNs, GNNs, and transformer-based LLMs. However, none of this content is present in the submitted full text. As a result, the significance cannot be assessed: the central theoretical and empirical support is entirely absent, and the submitted body is a different paper on cosmic-ray detection. The work therefore provides no basis for evaluation or acceptance.","major_comments":[{"comment":"The abstract describes a continual-learning theory with a new quantity CLEMC, a difference equation for capacity dynamics, and experiments across architectures. The full text is an AugerPrime/ICRC2025 astroparticle detector paper. There is no mention of CLEMC, continual learning, neural networks, capacity, stability-plasticity, or any related equations. This is a complete mismatch, so every load-bearing claim in the abstract is unsupported by the submitted manuscript.","section":"Abstract vs. full text"},{"comment":"The central derivation promised in the abstract—the difference equation modeling the interplay between NN, task data, and optimization—is nowhere in the manuscript. No mathematical definition of CLEMC is given, no assumptions or regime of validity are stated, and no proof is supplied. Consequently, the claim that the stability-plasticity balance point is 'inherently non-stationary' cannot be checked.","section":"Full text (all sections)"},{"comment":"The abstract claims 'extensive experiments' across MLPs, CNNs, GNNs, and transformer-based LLMs. The submitted full text contains no experimental results on any neural network architecture. The empirical support for the universal, architecture-independent claim is therefore completely absent. The claim may be true or false, but there is no evidence in this manuscript.","section":"Full text (all sections)"}],"minor_comments":[{"comment":"The arXiv metadata (title, abstract, and identifier) does not match the body text. If this is a submission error, the authors need to resubmit with the correct full text. If not, the provenance of the abstract and body needs clarification.","section":"General"}],"recommendation":"reject","confidential_remarks":"To the editor: The submitted full text is an astroparticle physics paper (AugerPrime status report) bearing no relation to the abstract's continual-learning claims. This is not a case of incomplete evidence or under-specification; it is a different paper. The central claims are entirely unsupported. I see no route to revision within this submission; the authors would need to supply the actual manuscript content. I recommend rejection of this submission as-is."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: the abstract describes a theory of continual learning capacity (CLEMC) and a difference equation, but the full text is an AugerPrime cosmic-ray detector paper. None of the promised material is present—no definitions, no equation, no experiments, no references to continual learning. As submitted, the paper is internally incoherent.\n\nThe abstract itself is not worthless. The idea of a formal effective capacity that evolves and explains stability-plasticity tradeoffs is worth taking seriously, and the empirical scope advertised (MLPs to LLMs) would be a useful testbed if it existed. So the authors may have a real project underneath. But the claim to a universal, architecture-independent law is exactly the kind of thing that needs a careful derivation and honest validation, and there is none here.\n\nThe soft spot is not a soft spot; it is the load-bearing wall missing. The difference equation is not shown, so we cannot tell whether the conclusion follows from the model or is baked into its assumptions. The experiments are not shown, so the empirical support is absent. And the full text being a different paper entirely means no amount of goodwill can fix it. This is not a subtle under-specification.\n\nI agree with the stress-test note: this is a mismatch between abstract and body, not a technical flaw to be patched. The citation pattern cannot be assessed because there are no citations in the body. There is no formal verification, no code, no data.\n\nMy recommendation: this should be desk rejected, or sent back to the authors to resubmit the correct manuscript. If the correct paper exists, it might deserve peer review. But the submission in front of us does not. No reader gets value from this as submitted; the abstract alone might interest CL researchers, but there is nothing to evaluate.","headline":"The abstract promises a continual-learning theory, but the full text is an AugerPrime detector paper—the submitted manuscript contains none of the claimed content.","tokens_in":3058,"tokens_out":3708,"would_cite":false,"duration_ms":37468,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper's abstract claims effective capacity in continual learning is non-stationary, making forgetting inevitable when task distributions shift; the body does not contain the promised derivation.","keywords":["continual learning","catastrophic forgetting","effective model capacity","stability-plasticity dilemma","non-stationarity","difference equation","neural networks","task distributions"],"falsifier":"A direct falsifier would be a continual-learning run—say, a transformer trained on a sequence of tasks with deliberately non-overlapping distributions—where a direct measure of effective capacity for the new task is measured to increase or stay flat rather than diminish; any such counterexample within the claimed architecture/optimizer scope would disprove the universality statement. More immediately, locating the promised difference equation in a complete manuscript and showing whether it admits non-decaying solutions for some task orderings would settle whether the theorem is true or an arti","tokens_in":2274,"feed_emoji":"🧠","tokens_out":9727,"duration_ms":90117,"temperature":0.7,"pith_summary":"The abstract of this submission proposes a quantity called CL's effective model capacity (CLEMC) to track the dynamic stability-plasticity balance point of a neural network trained on a sequence of tasks. It claims to derive a difference equation governing the interplay between the network, the task data, and the optimizer, and from that equation concludes that the effective capacity—and the balance point—is inherently non-stationary. The paper asserts that, regardless of architecture or optimization method, a network's ability to represent new tasks diminishes whenever incoming task distributions differ from previous ones. The body of the submission, however, is not the continual-learning paper at all: it is a status report on a cosmic-ray observatory's detector upgrade, with no equation, derivation, or experiment related to CLEMC. A sympathetic reader can only take the abstract's claims as the intended contribution, since the supporting text is absent.","feed_headline":"Forgetting called inevitable when tasks diverge; proof text is missing","feed_subtitle":"Abstract says capacity is non-stationary; the manuscript has no equation or experiments.","key_machinery":"The central object is CLEMC (CL's effective model capacity), a proposed quantity meant to characterize the instantaneous stability-plasticity balance point of a continual learner. The carrying mechanism is a first-order difference equation that supposedly describes how the network's effective capacity evolves as a function of the network, the task data, and the optimization procedure. The equation is the load-bearing formalism; without it, the non-stationarity conclusion has no demonstrated route. In the submitted text, this equation appears nowhere.","core_discovery":"On the paper's own terms, the central discovery is that the effective capacity of a neural network in a continual learning setting is a moving target: the stability-plasticity balance point shifts as tasks arrive, and this shift is governed by a difference equation coupling the network state, the task distribution, and the optimizer. From this, the author derives the universality claim: no architecture or optimization method can prevent the decline in representational ability for new tasks when those tasks come from distributions different from the training history. This would amount to a general law of catastrophic forgetting. The manuscript as submitted does not actually demonstrate this:","pith_inferences":["If the non-stationarity claim is true and universal, the natural next step is to measure CLEMC directly in controlled task sequences with varying distributional overlap; this would turn the abstract's law into a quantitative prediction about the rate of capacity decay.","The absence of the equation in the submitted manuscript means the present version does not allow a reader to distinguish a genuine derivation from a tautology forced by the model's definition; a revised version would need to state the equation and its regime of validity.","One could connect this to existing continual-learning theory by asking whether the difference equation reduces to known scaling laws in the limit of many tasks, or whether it predicts phase transitions in forgetting as distribution shift crosses a threshold.","Another extension: if the balance point is intrinsically non-stationary, then meta-learning or adaptive regularization that re-estimates capacity online should outperform fixed regularizers; that is a testable experimental prediction derived from the abstract's claim, not from the (missing) proof."],"forward_implications":["If CLEMC is correct, continual-learning systems cannot rely on a fixed capacity budget; the balance point itself drifts, so any static allocation of resources will be misaligned over time.","The claimed architecture-independence implies that the diminishing ability to represent new tasks is a property of the learning dynamics, not of a particular model family, so efforts to avoid forgetting must target the update rule or the task distribution rather than the network size.","A direct corollary is that the stability-plasticity tradeoff is not a single tunable hyperparameter but a trajectory; comparisons between CL methods should therefore be made over time, not at a single checkpoint.","The paper also implies that task-order matters in a specific way: the diminishment is tied to the difference between incoming and previous task distributions, so overlapping or gradually shifting tasks should cause less capacity loss."],"supporting_citations":[],"fun_headline_variants":["Capacity shift claimed universal, but equations absent","No architecture escapes forgetting? Paper lacks proof","Stability-plasticity balance called non-stationary, no evidence","Task divergence kills capacity, but manuscript omits math","Forgetting inevitability asserted, derivation missing"],"cache_read_input_tokens":2816,"weakest_assumption_plain":"The conclusion rests on a difference equation that the paper never presents, so the entire argument depends on the unstated premise that such an equation faithfully captures the coupled evolution of network, data, and optimizer; without seeing the equation, the claimed non-stationarity could be an artifact of the model's construction.","fun_headline_variants_meta":{"raw":{"variants":["Capacity shift claimed universal, but equations absent","No architecture escapes forgetting? Paper lacks proof","Stability-plasticity balance called non-stationary, no evidence","Task divergence kills capacity, but manuscript omits math","Forgetting inevitability asserted, derivation missing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000178,"raw_usage":{"total_tokens":1091,"prompt_tokens":662,"completion_tokens":429,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":406,"completion_tokens_details":{"reasoning_tokens":355}},"tokens_in":406,"tokens_out":429,"duration_ms":4816,"temperature":1.0,"reasoning_tokens":355,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T21:39:49.515087+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct falsifier would be a continual-learning run—say, a transformer trained on a sequence of tasks with deliberately non-overlapping distributions—where a direct measure of effective capacity for the new task is measured to increase or stay flat rather than diminish; any such counterexample within the claimed architecture/optimizer scope would disprove the universality statement. More immediately, locating the promised difference equation in a complete manuscript and showing whether it admits non-decaying solutions for some task orderings would settle whether the theorem is true or an arti","supporting_citations":[],"review_version":1}