{"id":"03136ba0-f472-4b35-be0e-eaf89db90647","arxiv_id":"2605.13297","paper_version":1,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"PaMM augments an equivariant atomistic model with explicit pair and triplet motif memory tables, producing modest gains in energy and force MAE on OMAT benchmarks at fixed training budget.","lead":"PaMM adds explicit hashed lookup tables for common pair and triplet local coordination motifs to an existing equivariant atomistic model. In controlled tests on OMAT data at fixed training steps, the added memory tables improve energy and force accuracy over the baseline without simply adding generic capacity.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Gains shown only at fixed 10k/20k-step budgets; no convergence curves or run-to-run variance reported","rationale":"The reader's weakest_assumption directly identifies the same load-bearing point (gains at intermediate checkpoints vs. capacity/optimization effects and persistence at convergence). The paper's narrow framing around a fixed training budget makes the claim internally consistent, but the absence of convergence data and variance estimates keeps the evidence strength modest, supporting the existing CONDITIONAL verdict with low confidence.","tokens_in":1844,"tokens_out":346,"duration_ms":25249,"concrete_test":"Re-train both baseline UMA-S and the two PaMM variants from scratch on the identical OMAT split until validation energy/force loss plateaus (or a fixed large step count, e.g. 100k+), reporting mean and std over at least three seeds; if the final MAE gap shrinks below the observed intermediate gap or loses statistical significance, the inductive-bias interpretation at fixed budget does not extend to the usual training regime.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that explicit pair/triplet motif memory supplies a useful inductive bias in the UMA-S + OMAT regime. This rests on the observed MAE improvements at the 10k- and 20k-step checkpoints plus the ablation controls (pair-only, triplet-only, random-bucket, parameter-matched MLP). Because training is halted at these intermediate points and no standard deviations across random seeds are supplied, it remains possible that the modest gains are transient optimization artifacts or lie inside normal run-to-run fluctuation rather than evidence of a stable structural advantage that would survive full convergence.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper introduces PaMM, a periodic motif memory module that augments the UMA eSCN-MD edge encoder with explicit hashed lookup tables for pair motifs keyed by (Z_j, Z_i, b_r) and triplet motifs keyed by (Z_j, Z_i, Z_k, b_θ). It evaluates this addition in a matched UMA-S + OMAT setting at fixed 10k- and 20k-step training budgets, reporting MAE gains over the plain baseline as well as pair-only, triplet-only, random-bucket, and parameter-matched MLP controls, and concludes that explicit pair/triplet motif memory supplies a useful inductive bias for periodic atomistic modeling in this regime.","tokens_in":1962,"tokens_out":458,"duration_ms":53869,"significance":"If the gains prove robust, the work supplies targeted empirical evidence that explicit local-structure memory can serve as an effective inductive bias in equivariant atomistic models, improving performance at intermediate training budgets without altering the core architecture. The use of matched controls and the narrow, falsifiable scope of the claim (fixed-budget UMA-S + OMAT) are strengths that would make the result a useful reference point for future architecture design in materials ML.","major_comments":[{"comment":"Evaluation at fixed checkpoints: The MAE improvements at the 10k- and 20k-step checkpoints are reported without standard deviations, multiple random seeds, or error bars. Given that the gains are described as modest, this omission prevents assessment of whether they exceed normal run-to-run fluctuation (abstract and results sections).","section":"Results"},{"comment":"Training dynamics: No learning curves to full convergence are provided, so it is unclear whether the observed advantage of the motif memory persists beyond the intermediate checkpoints or is specific to early-stage optimization behavior (experimental setup and evaluation sections).","section":"Experimental Setup"}],"minor_comments":[{"comment":"The description of the gate-only versus affine-equipped fusion variants would be clearer with an explicit equation or small diagram reference in the methods.","section":"Methods"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive comments on our manuscript. We address each major point below, indicating where revisions will be made.","responses":[{"response":"We agree that the absence of error bars from multiple seeds limits assessment of the modest gains. In the revised manuscript we will rerun the 10k- and 20k-step evaluations with three independent random seeds and report mean MAE values together with standard deviations.","revision_made":"yes","referee_comment":"[Results] Evaluation at fixed checkpoints: The MAE improvements at the 10k- and 20k-step checkpoints are reported without standard deviations, multiple random seeds, or error bars. Given that the gains are described as modest, this omission prevents assessment of whether they exceed normal run-to-run fluctuation (abstract and results sections)."},{"response":"The manuscript's scope is deliberately restricted to fixed intermediate training budgets, as stated in the abstract and introduction; we make no claim about behavior at full convergence. To address the request we will add a clarifying sentence in the evaluation section and include full learning curves (to 100k steps) for the main PaMM variants versus baseline in the supplementary material.","revision_made":"partial","referee_comment":"[Experimental Setup] Training dynamics: No learning curves to full convergence are provided, so it is unclear whether the observed advantage of the motif memory persists beyond the intermediate checkpoints or is specific to early-stage optimization behavior (experimental setup and evaluation sections)."}],"tokens_in":1496,"tokens_out":331,"duration_ms":26748,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"PaMM introduces explicit periodic motif memory by hashing pair motifs on (Z_j, Z_i, b_r) and triplet motifs on (Z_j, Z_i, Z_k, b_theta) into fixed tables, then fusing them into the UMA edge encoder via gate-only or affine layers. In the matched UMA-S OMAT setting the two variants beat the plain baseline on energy and force MAE at both the 10k and 20k checkpoints, and the structured ablations (pair-only, triplet-only, random-bucket, parameter-matched MLP) show weaker gains, pointing to the value of the element-and-geometry keyed organization rather than raw capacity.","headline":"PaMM adds explicit hashed pair/triplet motif memory to UMA-S and reports modest MAE gains at 10k/20k steps in OMAT, but the advantage rests on intermediate checkpoints without variance or full curves.","tokens_in":2484,"tokens_out":224,"would_cite":false,"duration_ms":25204,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":{"model":"grok-4.3","evidence":[],"headline":"PaMM motif hashing and gated fusion is orthogonal to RS forcing","alignment":"orthogonal","rationale":"The paper augments an equivariant edge encoder with fixed-size hashed lookup tables for recurring (Z_i,Z_j,b_r) pair and (Z_i,Z_j,Z_k,b_θ) triplet motifs, fused via scalar gates or per-layer affine modulation. This is a standard ML capacity-control technique for periodic data; it invokes none of the RS primitives (J-cost functional equation, φ-ladder, 8-tick periodicity, ratio-symmetric cost, or the distinction-to-spacetime forcing chain). No theorem from Cost/FunctionalEquation, Foundation/RealityFromDistinction, or AlexanderDuality is paralleled or contradicted.","tokens_in":51034,"confidence":"high","tokens_out":169,"duration_ms":12582,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Explicit pair and triplet motif lookup tables improve energy and force accuracy in periodic atomistic models at intermediate training steps.","keywords":["periodic motifs","atomistic modeling","equivariant networks","local structure","inductive bias","crystal prediction","edge encoding","machine learning potentials"],"falsifier":"Full-convergence training runs in which the PaMM variants show no final MAE advantage over the baseline, or in which random-bucket controls match the structured-motif performance, would falsify the claim that explicit motif memory supplies a useful inductive bias.","tokens_in":2697,"feed_emoji":"⚛️","tokens_out":845,"duration_ms":35862,"temperature":0.7,"pith_summary":"Current equivariant models for crystals encode repeating local coordination patterns only implicitly through dense edge features. PaMM adds explicit memory tables that store pair motifs keyed by element types and radial bins and triplet motifs keyed by element types and angular bins, then fuses those lookups back into the edge representation. In controlled runs on the OMAT dataset using the UMA-S architecture, both gate-only and affine-equipped versions of PaMM lower mean absolute errors for energy and forces at the 10k-step and 20k-step checkpoints. Ablations show that the gains shrink when the tables are replaced by random buckets or single-motif variants, indicating that the benefit comes from the structured organization of motifs rather than extra capacity alone. The result holds across held-out source families within the dataset, supporting the claim that explicit motif memory supplies a helpful inductive bias for periodic atomistic modeling under these training conditions.","feed_headline":"Motif lookup tables cut early errors in crystal models","feed_subtitle":"Explicit pair and triplet memory keyed on elements and geometry improves energy and force accuracy over baseline encoders at 10k steps.","key_machinery":"PaMM periodic motif memory that hashes pair motifs by (Z_j, Z_i, b_r) and triplet motifs by (Z_j, Z_i, Z_k, b_θ) into fixed-size tables and fuses them with edge features via gate or affine modules.","core_discovery":"PaMM augments the UMA eSCN-MD edge encoder with hashed lookup tables for pair motifs keyed by (Z_j, Z_i, b_r) and triplet motifs keyed by (Z_j, Z_i, Z_k, b_θ). These tables are fused with the baseline edge features through lightweight gate-only or affine-equipped modules. In matched UMA-S + OMAT experiments, the gate-only variant records the lowest energy MAE and the affine variant the lowest force MAE at both 10k and 20k steps, while pair-only, triplet-only, random-bucket, and capacity-matched MLP controls produce smaller gains. Within-OMAT24 source-family splits likewise show small consistent improvements, establishing that explicit pair/triplet motif memory functions as a useful inductive","pith_inferences":["If the same pattern persists at full convergence on other datasets, explicit motif memory could lower the total training compute needed for periodic systems.","The approach may transfer to other equivariant architectures that currently rely on implicit edge features.","Inspectable motif tables could support post-hoc analysis of which local geometries the model treats as similar across different crystals."],"forward_implications":["At fixed intermediate budgets of 10k and 20k steps, both PaMM variants outperform the plain UMA-S baseline on energy and force MAEs.","Combined pair-plus-triplet tables produce larger gains than pair-only, triplet-only, or random-bucket alternatives.","Small but consistent improvements appear across held-out generation families within the OMAT24 source split.","The motif tables provide an inspectable local-structure interface that remains compatible with the existing equivariant encoder."],"fun_headline_variants":["PaMM adds hashed motif tables to atomistic crystal encoders","Explicit pair triplet memory improves early training in crystal models","Motif lookups keyed by elements and bonds aid periodic structures","PaMM variants record lower MAE than plain UMA baseline at 10k steps"],"cache_read_input_tokens":64,"weakest_assumption_plain":"The measured gains at the 10k-step and 20k-step checkpoints are produced by the structured motif tables rather than by incidental capacity increases or optimization differences.","fun_headline_variants_meta":{"raw":{"variants":["PaMM adds hashed motif tables to atomistic crystal encoders","Explicit pair triplet memory improves early training in crystal models","Motif lookups keyed by elements and bonds aid periodic structures","PaMM variants record lower MAE than plain UMA baseline at 10k steps"]},"model":"grok-4.3","cost_usd":0.006914,"raw_usage":{"total_tokens":3213,"prompt_tokens":841,"num_sources_used":0,"completion_tokens":69,"cost_in_usd_ticks":69140500,"prompt_tokens_details":{"text_tokens":841,"audio_tokens":0,"image_tokens":0,"cached_tokens":64},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2303,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":841,"tokens_out":69,"duration_ms":37297,"temperature":1.0,"reasoning_tokens":2303,"cache_read_input_tokens":64,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-05-14T19:28:35.463373+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Full-convergence training runs in which the PaMM variants show no final MAE advantage over the baseline, or in which random-bucket controls match the structured-motif performance, would falsify the claim that explicit motif memory supplies a useful inductive bias.","supporting_citations":[],"review_version":1}