{"id":"8d3dbacd-c8ed-4d14-9a8e-113c6f5caec2","arxiv_id":"2606.29679","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Observable Matrix Dynamics (OMD) is a new diagnostic framework that uses random matrix theory on distance matrices to distinguish diffusive relaxations from phase-transition-like reorganizations during neural network training.","lead":"The paper introduces Observable Matrix Dynamics (OMD), a diagnostic that tracks neural network training by monitoring spectral changes in a fixed-size distance matrix of held-out input representations. A smart generalist might read it because it offers a geometric view of learning regimes that standard loss curves overlook, potentially aiding better training diagnostics.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Applicability of BBS perturbative ambient-latent decomposition to NN distance matrices M(t) remains unverified","rationale":"The reader’s weakest_assumption is exactly the load-bearing step. The seven experiments show empirical patterns, but without an explicit check that the matrix ensemble obeys the BBS assumptions, the geometric interpretation cannot be secured. This keeps the verdict CONDITIONAL pending the synthetic-matrix control.","tokens_in":1735,"tokens_out":312,"duration_ms":15222,"concrete_test":"Construct synthetic N×N distance matrices from points uniformly sampled on a 2-torus and on a product of two circles (matching the N and approximate ambient dimension of the paper’s experiments), add controlled ambient noise at the levels reported for the NN runs, and compute the top-10 eigenvalues; if the resulting band structure deviates from the perturbative BBS predictions used to interpret the NN M(t), the decomposition does not transfer.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The strongest claim—that OMD distinguishes geometric regimes via stable top-of-spectrum fingerprints—rests on the perturbative extension of BBS random-matrix theory applying to the extracted M(t). M(t) is a deterministic Euclidean distance matrix on learned representations, not an ensemble of random distance matrices on a manifold; the paper supplies no derivation showing that the eigenvalue repulsion or band-structure corrections survive the non-random, input-correlated structure of NN embeddings. If the decomposition fails here, the mapping from observed band stability to “smooth or product latent geometries” is unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper introduces Observable Matrix Dynamics (OMD), a framework that constructs a fixed-size N×N Euclidean distance matrix M(t) from a held-out set of inputs at each training snapshot t of a neural network, then applies random-matrix diagnostics (top-of-spectrum band structure, ambient noise) and a 3D MDS embedding to track spectral reorganizations. It extends the Bogomolny–Bohigas–Schmit (BBS) perturbative ambient-versus-latent decomposition to these matrices and reports that diffusive regimes lack stable band structure while sharp endogenous or externally driven reorganizations produce stable fingerprints consistent with smooth/product latent geometries (or finite-cluster/Fourier-soliton structures). The central claim is that OMD reads the geometric regime of a representation rather than a single intrinsic dimension, demonstrated across seven experiments.","tokens_in":1843,"tokens_out":483,"duration_ms":17548,"significance":"If the BBS extension is shown to apply, OMD would supply a geometrically interpretable, falsifiable diagnostic that distinguishes training regimes missed by scalar losses and links observed spectral stability to latent geometry classes; the provision of trajectory-level observables and reproducible MDS visualizations would be a concrete strength.","major_comments":[{"comment":"The central claim—that stable top-of-spectrum fingerprints map to smooth or product latent geometries—rests on the perturbative ambient-versus-latent decomposition of BBS theory applying to the deterministic Euclidean distance matrices M(t) extracted from NN representations. No derivation is supplied showing that eigenvalue repulsion or band-structure corrections survive the non-random, input-correlated structure of these matrices (as opposed to ensembles of random distance matrices on a manifold).","section":"Abstract and the section introducing the BBS extension"},{"comment":"The abstract states experimental outcomes across seven experiments (diffusive regimes lack stable band structure; sharp reorganizations produce stable fingerprints) but supplies no derivations, error analysis, data details, or validation steps for the per-snapshot diagnostics or the mapping from observed band stability to geometry classes.","section":"Abstract"}],"minor_comments":[{"comment":"Notation for the distance matrix M(t) and the precise definition of the top-of-spectrum band should be introduced with an equation number on first use.","section":null}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive and detailed report. We address each major comment below, indicating planned revisions where appropriate.","responses":[{"response":"We acknowledge that the manuscript does not contain a full analytic derivation establishing that the BBS perturbative decomposition (eigenvalue repulsion and band-structure corrections) carries over exactly to deterministic Euclidean distance matrices M(t) whose entries are correlated through the neural-network representation. The extension is motivated by the fact that each M(t) remains a Euclidean distance matrix on the representation manifold, and the observed diagnostics are presented as an empirical extension of BBS rather than a proven identity. In the revised manuscript we will add an explicit subsection in the methods clarifying the assumptions, the heuristic character of the extension, and the empirical support from the experiments; we will also note that a complete proof under input correlations lies outside the present scope.","revision_made":"partial","referee_comment":"[Abstract and the section introducing the BBS extension] The central claim—that stable top-of-spectrum fingerprints map to smooth or product latent geometries—rests on the perturbative ambient-versus-latent decomposition of BBS theory applying to the deterministic Euclidean distance matrices M(t) extracted from NN representations. No derivation is supplied showing that eigenvalue repulsion or band-structure corrections survive the non-random, input-correlated structure of these matrices (as opposed to ensembles of random distance matrices on a manifold)."},{"response":"The abstract is written as a concise summary of the central findings; the derivations, error analysis, data specifications, and validation procedures for the per-snapshot diagnostics and geometry-class mapping are supplied in Sections 2–4 and the appendices. To improve clarity we will revise the abstract to include a single sentence directing readers to those sections for the technical details of the diagnostics and the empirical mapping.","revision_made":"yes","referee_comment":"[Abstract] The abstract states experimental outcomes across seven experiments (diffusive regimes lack stable band structure; sharp reorganizations produce stable fingerprints) but supplies no derivations, error analysis, data details, or validation steps for the per-snapshot diagnostics or the mapping from observed band stability to geometry classes."}],"tokens_in":1417,"tokens_out":423,"duration_ms":26013,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main takeaway is that this paper introduces Observable Matrix Dynamics as a way to watch how a neural network's internal representations evolve by tracking a fixed N by N distance matrix over training steps. It applies random matrix ideas, specifically an extension of Bogomolny-Bohigas-Schmit theory, to spot diffusive regimes versus sharp reorganizations through band structure and a 3D MDS embedding of the trajectory.\n\nWhat stands out as new is the set of trajectory-level observables and the 3D visualization that treat training as particle motion in a reduced space. The seven experiments illustrate that diffusive phases show unstable top-of-spectrum features while reorganizations produce more stable patterns, which scalar losses do not capture. This gives a geometric reading of the process rather than a single dimension number.\n\nThe soft spot is the central assumption. The stress-test note correctly flags that the perturbative ambient-versus-latent decomposition was developed for random distance matrices on manifolds, yet here M(t) comes from deterministic, input-correlated embeddings. The abstract supplies no derivation showing the eigenvalue repulsion or band corrections survive this change, and no error analysis or controls appear. Without those steps the mapping from observed fingerprints to smooth or product geometries stays unsupported.\n\nThe work targets readers who study training dynamics beyond loss curves, such as people in representation learning or dynamical systems approaches to ML. It deserves a serious referee because the diagnostic idea is coherent and the experiments are described, even if the theoretical link needs checking. I would not cite it until the BBS applicability is shown explicitly.","headline":"OMD tracks representation changes via distance matrix spectra but the BBS extension to NN embeddings lacks verification in the given material.","tokens_in":2310,"tokens_out":375,"would_cite":false,"duration_ms":19515,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Observable Matrix Dynamics distinguishes diffusive relaxation from sharp geometric reorganizations in neural network training via fixed-size distance matrices.","keywords":["observable matrix dynamics","random matrix theory","neural network representations","distance matrix","spectral analysis","learning dynamics","phase transitions"],"falsifier":"A controlled experiment in which a sharp reorganization of representations occurs yet the top-of-spectrum band structure remains unstable or absent would falsify the claim that reorganizations produce stable geometric fingerprints.","tokens_in":2621,"feed_emoji":"📐","tokens_out":815,"duration_ms":23064,"temperature":0.7,"pith_summary":"The paper introduces Observable Matrix Dynamics as a way to track how neural networks reorganize their internal representations during training. It extracts a fixed N by N distance matrix from a held-out set of inputs at each step and applies tools from random matrix theory to monitor spectral changes that scalar losses overlook. Experiments across seven settings show that smooth diffusive regimes produce no stable top-of-spectrum structure, while both endogenous and externally triggered reorganizations leave consistent fingerprints that match expected signatures of smooth, product, cluster, or soliton geometries. The method therefore reads the geometric regime of a representation rather than collapsing it to one intrinsic-dimension number.","feed_headline":"Distance matrices reveal diffusion versus phase transitions in neural training","feed_subtitle":"Fixed-size M(t) snapshots distinguish smooth relaxation from abrupt geometric reorganizations missed by loss curves.","key_machinery":"Observable Matrix Dynamics (OMD) applied to the time-evolving distance matrix M(t), read through a perturbative ambient-versus-latent decomposition extending BBS random-matrix theory, with top-of-spectrum band diagnostics and 3D MDS trajectory embeddings.","core_discovery":"Observable Matrix Dynamics (OMD) is a diagnostic framework that probes the dynamics of high-dimensional internal representations of inputs by a neural network via a fixed-size N × N distance matrix M(t) on a held set of N inputs. OMD uses methods of random matrix theory and particle dynamics to explore spectral reorganisations that are missed by scalar loss functions, but are informative of the training process. We read M(t) against a perturbative ambient-versus-latent decomposition extending the Bogomolny--Bohigas--Schmit (BBS) theory of random distance matrices, with per-snapshot diagnostics for the top-of-spectrum band structure and ambient noise, trajectory-level observables linking snap","pith_inferences":["OMD could be applied to detect phase-transition-like events in other high-dimensional dynamical systems whose state is captured by evolving distance matrices.","The method supplies a concrete diagnostic for when a representation has settled into a stable latent geometry versus continued diffusion.","Because it operates on a fixed held-out set, OMD can be inserted into existing training loops without changing the optimization itself."],"forward_implications":["Diffusive training regimes are diagnosed by the absence of persistent top-of-spectrum band structure in M(t).","Sharp endogenous or externally driven reorganizations leave stable fingerprints whose geometry can be classified as smooth, product, cluster, or soliton type.","Scalar loss curves miss the spectral reorganizations that OMD detects at the level of the representation geometry.","Training trajectories can be visualized as a moving particle cloud in the bottom-three eigenvectors of M(t)."],"fun_headline_variants":["Matrix dynamics distinguish diffusion from phase transitions","Distance matrices track diffusive relaxations in training","Spectral reorganizations in neural representations via M(t)","Observable matrix dynamics read training geometric regimes"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The perturbative ambient-versus-latent decomposition extending BBS theory of random distance matrices applies to the distance matrices extracted from neural network internal representations.","fun_headline_variants_meta":{"raw":{"variants":["Matrix dynamics distinguish diffusion from phase transitions","Distance matrices track diffusive relaxations in training","Spectral reorganizations in neural representations via M(t)","Observable matrix dynamics read training geometric regimes"]},"model":"grok-4.3","cost_usd":0.007183,"raw_usage":{"total_tokens":3256,"prompt_tokens":712,"num_sources_used":0,"completion_tokens":53,"cost_in_usd_ticks":71828000,"prompt_tokens_details":{"text_tokens":712,"audio_tokens":0,"image_tokens":0,"cached_tokens":64},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2491,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":712,"tokens_out":53,"duration_ms":30352,"temperature":1.0,"reasoning_tokens":2491,"cache_read_input_tokens":64,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-30T06:58:34.195386+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A controlled experiment in which a sharp reorganization of representations occurs yet the top-of-spectrum band structure remains unstable or absent would falsify the claim that reorganizations produce stable geometric fingerprints.","supporting_citations":[],"review_version":1}