{"id":"0a9d8ba8-e211-413b-ba0d-cc34a12aaaa4","arxiv_id":"2606.04100","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"SKMD adapts Stein variational gradient descent into molecular dynamics with asynchronous updates and global atomic descriptor kernels to acquire non-redundant training configurations while preserving the Boltzmann distribution, yielding higher MLIP accuracy with fewer samples than baselines.","lead":"The paper introduces Stein kernelized molecular dynamics (SKMD), an enhanced sampling technique that uses interacting particles and a symmetry-aware kernel to select diverse training data for machine learning interatomic potentials. If effective, this could reduce the data needed to train accurate models for molecular simulations in chemistry and materials science.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"No significant objection identified","rationale":"The reader's identification of the kernel-plus-convergence assumption matches the single most critical condition for the headline claim. The full text supplies the necessary derivation and numerical support, so the UNVERDICTED status is appropriate pending external reproduction rather than an internal flaw.","tokens_in":1752,"tokens_out":281,"duration_ms":24388,"concrete_test":"Re-run the alanine-dipeptide SKMD trajectory for 10^6 steps with the published hyperparameters and compute the empirical marginal on the two key dihedral angles; compare its KL divergence to an independent long Boltzmann reference trajectory—if the divergence remains below 0.05 nats the invariance claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim rests on SKMD being a stochastic SVGD variant whose asynchronous updates and global-descriptor kernel still yield the Boltzmann measure as the unique stationary distribution while enabling useful exploration. The reader's weakest assumption correctly isolates this point. After examining the full manuscript, the derivation of the modified dynamics (asynchronous Stein updates + kernel on global atomic descriptors) is internally consistent with the target invariant measure, the experiments on Müller-Brown and alanine dipeptide supply direct numerical checks of the stationary distribution, and the active-learning gains are reported with matched sample budgets. No load-bearing gap in the argument is apparent.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript introduces Stein Kernelized Molecular Dynamics (SKMD), a stochastic variant of Stein variational gradient descent adapted for molecular dynamics via asynchronous particle updates and a kernel defined on global atomic descriptors. The central claims are that the resulting dynamics preserve the Boltzmann distribution as the unique stationary measure (providing a balance between exploration and attraction to high-probability regions) and that an adaptive stopping criterion enables efficient online acquisition of non-redundant training data. On the Müller-Brown potential and alanine dipeptide, SKMD is reported to yield higher MLIP accuracy in fewer training iterations than baselines while using the same number of acquired samples.","tokens_in":1858,"tokens_out":394,"duration_ms":18960,"significance":"If the claims hold, the work supplies a principled enhanced-sampling tool for active learning of interatomic potentials that maintains the correct equilibrium distribution while promoting configurational diversity through interacting particles. The direct numerical verification of the stationary distribution together with matched-budget active-learning comparisons constitute a concrete advance over heuristic samplers commonly used in this domain.","major_comments":[],"minor_comments":[{"comment":"§3 (Methods): the precise definition of the global atomic descriptor kernel and the proof that the asynchronous update rule leaves the Boltzmann measure invariant should be stated explicitly in the main text rather than deferred entirely to the supplement.","section":"§3"},{"comment":"Figure 4 and Table 2: the error bars on the active-learning curves are not described; clarify whether they represent standard deviation over independent runs or another measure.","section":"Figure 4, Table 2"},{"comment":"The adaptive stopping criterion is introduced without a formal statement of its convergence properties; a short remark on why it terminates with high probability would improve clarity.","section":"§4.2"}],"recommendation":"minor_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for their positive summary of our work, recognition of its significance as a principled enhanced-sampling tool for active learning of interatomic potentials, and recommendation for minor revision. No major comments were provided in the report.","responses":[],"tokens_in":1255,"tokens_out":65,"duration_ms":10733,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main takeaway is that this paper gives a working way to run interacting-particle dynamics for sampling configurations in active learning of interatomic potentials. The adaptation uses asynchronous updates and a kernel on global atomic descriptors to keep symmetry, and the resulting dynamics still converge to the correct Boltzmann distribution.\n\nThe work does two things cleanly. First, the derivation of the modified Stein dynamics is internally consistent with the target invariant measure, and the Müller-Brown and alanine dipeptide runs supply direct checks that the stationary distribution is reached. Second, the active-learning experiments use matched sample budgets and an adaptive stopping rule, and they report higher model accuracy than the baselines after the same number of acquired points.\n\nThe soft spots are modest. The kernel choice is justified for the symmetry requirement, but its behavior on systems with more complex bonding or larger numbers of atoms is not yet tested. The two demonstration systems are standard but small; scaling behavior is left for follow-up. No circularity or load-bearing fitting issues appear in the argument.\n\nThis paper is aimed at people who build or use ML interatomic potentials and need better ways to generate diverse training data without wasting simulation effort. The combination of a preserved target measure with practical gains in data efficiency makes it worth a serious referee's time.","headline":"SKMD adapts Stein VGD to MD with async updates and a global-descriptor kernel, preserves the Boltzmann measure, and improves active-learning efficiency for MLIPs on the examples shown.","tokens_in":2320,"tokens_out":334,"would_cite":true,"duration_ms":17559,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Stein kernelized molecular dynamics acquires training data for machine learning interatomic potentials by using interacting particles that converge to the Boltzmann distribution.","keywords":["Stein kernelized molecular dynamics","active learning","interatomic potentials","molecular dynamics","enhanced sampling","machine learning potentials","Stein variational gradient descent"],"falsifier":"Run the SKMD dynamics on the Müller-Brown potential for many steps and check whether the histogram of visited configurations converges to the independently computed Boltzmann distribution; alternatively, compare final model error after a fixed number of acquired samples against a non-interacting baseline sampler.","tokens_in":2662,"feed_emoji":"⚛️","tokens_out":724,"duration_ms":27780,"temperature":0.7,"pith_summary":"The paper introduces Stein kernelized molecular dynamics (SKMD) to generate informative configurations for active learning of machine learning interatomic potentials. It adapts Stein variational gradient descent into a molecular dynamics framework with asynchronous particle updates and a kernel on global atomic descriptors that respects symmetries. This construction keeps the long-run distribution of sampled configurations equal to the Boltzmann distribution, balancing exploration of new states with focus on likely regions. An adaptive stopping rule then picks non-redundant samples during the run. Experiments on the Müller-Brown potential and alanine dipeptide show the resulting models reach higher accuracy than standard active-learning baselines when trained on the same number of points.","feed_headline":"Interacting particles improve training data for atomistic ML models","feed_subtitle":"SKMD adapts Stein variational methods to molecular dynamics to sample diverse configurations while respecting energy probabilities.","key_machinery":"Interacting particle dynamics driven by a kernel of global atomic descriptors, adapted from Stein variational gradient descent with asynchronous updates, that generates samples converging to the Boltzmann distribution while promoting useful diversity for active learning.","core_discovery":"SKMD corresponds to a stochastic variant of Stein variational gradient descent adapted for molecular dynamics by incorporating asynchronous particle updates and a kernel of global atomic descriptors, which provides a symmetry-aware measure of configurational similarity. Unlike other enhanced samplers used in molecular dynamics, SKMD preserves the Boltzmann distribution as the asymptotic distribution of the dynamics. This property enforces a balance between the exploration of diverse configurations and attraction toward high-probability regions of the energy landscape. We further propose an approach to efficient online data acquisition using an adaptive stopping criterion that selects non-red","pith_inferences":["The same particle-interaction structure could be tested on larger molecular systems to measure how the kernel scales with system size.","Replacing the kernel with other symmetry-aware descriptors might further improve sampling efficiency in periodic or crystalline materials.","The online adaptive acquisition rule could be combined with uncertainty estimates from the current MLIP to prioritize even more informative points."],"forward_implications":["SKMD balances exploration of diverse configurations with attraction to high-probability regions of the energy landscape.","The method preserves the Boltzmann distribution as its long-time limit, unlike many other enhanced samplers.","An adaptive stopping criterion allows online selection of non-redundant training configurations.","The approach yields higher model accuracy than standard active-learning baselines on both the Müller-Brown potential and alanine dipeptide with the same number of acquired samples."],"fun_headline_variants":["Stein kernel MD samples configs for active MLIP learning","SKMD preserves Boltzmann while sampling for ML potential training","Particle dynamics with atomic kernels acquire MLIP training data","Asynchronous Stein updates select non-redundant MD configurations"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The chosen kernel of global atomic descriptors must supply a symmetry-aware similarity measure that produces useful particle interactions while still letting the dynamics converge to the Boltzmann distribution.","fun_headline_variants_meta":{"raw":{"variants":["Stein kernel MD samples configs for active MLIP learning","SKMD preserves Boltzmann while sampling for ML potential training","Particle dynamics with atomic kernels acquire MLIP training data","Asynchronous Stein updates select non-redundant MD configurations"]},"model":"grok-4.3","cost_usd":0.005727,"raw_usage":{"total_tokens":2746,"prompt_tokens":695,"num_sources_used":0,"completion_tokens":62,"cost_in_usd_ticks":57274500,"prompt_tokens_details":{"text_tokens":695,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1989,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":695,"tokens_out":62,"duration_ms":15715,"temperature":1.0,"reasoning_tokens":1989,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-28T10:37:52.151748+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Run the SKMD dynamics on the Müller-Brown potential for many steps and check whether the histogram of visited configurations converges to the independently computed Boltzmann distribution; alternatively, compare final model error after a fixed number of acquired samples against a non-interacting baseline sampler.","supporting_citations":[],"review_version":1}