{"id":"4b23cff7-9feb-44e0-b5d9-d904aef7ee8f","arxiv_id":"2605.22640","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Truncation to enforce positive-definiteness in separable priors for matrices distorts interpretability and biases sparse posterior inference unless off-diagonal variances are scaled with dimension.","lead":"This paper studies how adding a truncation step to enforce positive-definiteness in priors that treat matrix entries as independent can unintentionally change the prior's behavior. It shows that adjusting the variance of off-diagonal entries with matrix size can reduce distortion, especially avoiding extra bias toward sparse structures in sparse inference.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.3","headline":"Comparison to untruncated independent-entry distribution assumes it is a meaningful target, but it places mass outside the PD cone","rationale":"This is the same load-bearing assumption the reader already flagged. The full text would be required to check whether the authors explicitly defend the untruncated distribution as the right benchmark or instead compare only within the PD cone; absent that defense the claim remains conditional on accepting the comparison.","tokens_in":1628,"tokens_out":349,"duration_ms":40724,"concrete_test":"For the 3x3 case with independent N(0,σ²) entries (σ²=1), compute the probability that the (1,2) entry is exactly zero under (a) the untruncated distribution conditioned on the matrix being PD and (b) the truncated distribution; if the two probabilities differ by more than sampling error, the mass shift is not solely due to the PD constraint.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that truncation causes the prior (and posterior) to assign systematically higher mass to sparser structures than the untruncated version, unless parameters are tuned. This comparison treats the untruncated independent-entry distribution as the reference whose sparsity and shrinkage properties should be preserved. However, the untruncated distribution is defined on all symmetric matrices and assigns positive probability to non-PD matrices; its margins and sparsity characteristics are therefore not properties of a valid prior on the target space. Any observed difference in mass on sparse structures could be an artifact of projecting away the non-PD mass rather than an unintended side-effect of truncation per se. The paper's guidance on setting off-diagonal variances to mitigate the effect as dimension grows inherits this comparison.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript examines the effects of adding a truncation to independent-entry priors on symmetric matrices to enforce positive-definiteness. It claims that, unless prior parameters (especially off-diagonal variances) are chosen carefully, the truncated prior and resulting posterior assign systematically higher mass to sparser structures than the untruncated counterpart, for both dense and sparse settings; the paper provides guidance on parameter settings to mitigate this discrepancy as matrix dimension grows.","tokens_in":1745,"tokens_out":411,"duration_ms":26336,"significance":"If the claimed truncation effects and mitigation rules hold under rigorous derivation, the work would be significant for Bayesian covariance modeling and sparse precision-matrix inference, as it directly addresses interpretability and unintended shrinkage in a widely used prior class.","major_comments":[{"comment":"Abstract and introduction: the central claim rests on comparing the truncated prior to the untruncated independent-entry distribution as the reference whose sparsity properties should be preserved. However, the untruncated distribution is supported on all symmetric matrices and places positive mass outside the positive-definite cone; any observed difference in mass on sparse structures could therefore be an artifact of the projection onto the cone rather than an intrinsic effect of truncation. This comparison requires explicit justification or re-framing as a diagnostic rather than a normative target.","section":"Abstract"},{"comment":"The mitigation strategy of setting off-diagonal variances to control the effect as dimension grows inherits the same reference-distribution issue; without a clear statement of what properties of the untruncated margins are desirable on the PD cone, it is unclear whether the recommended parameter scaling achieves the intended preservation of interpretability.","section":"Abstract"}],"minor_comments":[{"comment":"The abstract states that the analysis covers both dense and sparse matrices, but does not indicate whether the mitigation rules differ between the two regimes or whether the same variance scaling applies.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for their constructive comments. We address the major comments point by point below.","responses":[{"response":"The untruncated independent-entry prior serves as the natural baseline for separable priors, with truncation applied subsequently to enforce positive-definiteness. Our analysis demonstrates the distortion introduced by this truncation. We have revised the manuscript to explicitly frame the comparison as a diagnostic for assessing truncation effects on interpretability and sparsity, rather than positioning the untruncated distribution as a normative target on the positive-definite cone. This clarification addresses the concern directly.","revision_made":"yes","referee_comment":"[Abstract] Abstract and introduction: the central claim rests on comparing the truncated prior to the untruncated independent-entry distribution as the reference whose sparsity properties should be preserved. However, the untruncated distribution is supported on all symmetric matrices and places positive mass outside the positive-definite cone; any observed difference in mass on sparse structures could therefore be an artifact of the projection onto the cone rather than an intrinsic effect of truncation. This comparison requires explicit justification or re-framing as a diagnostic rather than a normative target."},{"response":"We agree that the target properties on the PD cone merit explicit statement. The recommended scaling of off-diagonal variances is designed to ensure that the truncated prior's marginal distributions and sparsity characteristics more closely match those of the untruncated prior as dimension increases. We have added clarification in the revised manuscript specifying the desirable properties (matching marginal variances and reduced bias in sparsity) and demonstrating how the scaling achieves this approximation within the positive-definite cone.","revision_made":"yes","referee_comment":"[Abstract] The mitigation strategy of setting off-diagonal variances to control the effect as dimension grows inherits the same reference-distribution issue; without a clear statement of what properties of the untruncated margins are desirable on the PD cone, it is unclear whether the recommended parameter scaling achieves the intended preservation of interpretability."}],"tokens_in":1269,"tokens_out":418,"duration_ms":32553,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main thing to know is that this paper flags truncation effects in separable priors for positive definite matrices, but the central comparison to the untruncated version looks shaky on its face.\n\nThe work examines how adding a truncation step to enforce positive definiteness changes the prior and posterior relative to independent entries on all symmetric matrices. It covers both dense and sparse cases and derives rules for scaling the off-diagonal variances so the truncation impact fades as dimension grows. In the sparse inference section it argues that without those adjustments the truncated prior and posterior put extra mass on sparser structures.\n\nThis is a focused, practical check on a modeling choice that shows up often in multivariate Bayesian work. The dimension-dependent tuning advice is concrete and could help practitioners who already use these priors.\n\nThe soft spot is the reference point itself. The untruncated distribution assigns positive probability outside the positive definite cone, so its margins and sparsity profile are not properties of a valid prior on the target space. Any extra mass on sparse structures after truncation could simply reflect the removal of that invalid probability rather than an unintended side effect of the truncation mechanism. The paper's mitigation rules inherit the same comparison. If the derivations do not address this directly, the claim that truncation systematically favors sparsity needs rethinking.\n\nThe paper is for people who set or use separable priors in covariance estimation and graphical models. A reader who cares about prior interpretability in high dimensions will find the issue worth seeing, even if the fix is not yet settled.\n\nI would send it to peer review so referees can examine the technical details and decide whether the reference distribution choice holds up.","headline":"The truncation bias claim rests on comparing to an untruncated independent-entry distribution that isn't supported on positive definite matrices, so the sparsity effect may be an artifact rather than a real distortion.","tokens_in":2202,"tokens_out":406,"would_cite":false,"duration_ms":32291,"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":"Truncation to enforce positive definiteness in separable priors can systematically favor sparser matrices unless off-diagonal variances are adjusted with dimension.","keywords":["positive-definite priors","truncation effects","sparse inference","Bayesian covariance estimation","prior interpretability","separable priors","shrinkage priors"],"falsifier":"Compute or estimate the prior probability of matrices with a given number of zero off-diagonal entries under both the truncated and untruncated versions for fixed variance parameters as dimension increases; persistent deviation from equality for sparser cases would support the claim.","tokens_in":2522,"feed_emoji":"","tokens_out":654,"duration_ms":42968,"temperature":0.7,"pith_summary":"This paper examines how adding a truncation step to independent-entry priors for positive-definite matrices alters their properties compared to the untruncated version. The truncation ensures the matrix is positive definite but can change the distribution's mass on different sparsity levels. For sparse settings, this leads to the truncated prior and posterior putting more weight on sparser structures than intended. The authors show how to choose prior variances so that these differences diminish as the matrix size increases. A sympathetic reader would care because many Bayesian models for covariance matrices rely on such priors, and unintended shifts in sparsity can affect model selection and inference without the user realizing.","feed_headline":"Truncation in positive-definite priors biases to sparser matrices","feed_subtitle":"Scaling off-diagonal variances with dimension can prevent unintended shifts in mass assignment to sparse structures.","key_machinery":"The truncation operation applied to an independent-entry distribution to restrict support to the set of positive-definite matrices.","core_discovery":"The paper claims that for priors on symmetric positive-definite matrices that start with independent entries and then truncate to the positive-definite cone, the resulting distribution differs from the untruncated one in ways that affect interpretability and shrinkage. Specifically, unless the variance of off-diagonal entries is set to decrease appropriately with matrix dimension, the truncated prior assigns higher probability to sparser matrices, and this bias carries over to the posterior.","pith_inferences":["Users of these priors in high-dimensional settings may need explicit scaling rules in software defaults to avoid unintended sparsity bias.","Similar truncation adjustments could be required for other constrained matrix distributions such as correlation matrices.","Direct Monte Carlo comparison of truncated and untruncated samples in moderate dimensions would quantify the mass shift on sparsity levels."],"forward_implications":["Setting the variance of off-diagonal entries to scale with dimension mitigates the truncation effect for both dense and sparse matrices.","In sparse inference, careful parameter choice prevents the truncated prior and posterior from assigning systematically higher mass to sparser structures.","Posterior inference can be affected in unanticipated ways if truncation effects on mass assignment are ignored.","The shrinkage properties of the prior become harder to characterise without matching the untruncated margins."],"fun_headline_variants":["Truncation in separable priors causes sparsity bias","Scaling off-diagonal variances counters truncation bias","Truncated priors assign more mass to sparse matrices","Prior truncation affects sparsity in positive-definite inference"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The untruncated independent-entry distribution is the desired target that truncation should preserve as closely as possible for interpretability and shrinkage characterization.","fun_headline_variants_meta":{"raw":{"variants":["Truncation in separable priors causes sparsity bias","Scaling off-diagonal variances counters truncation bias","Truncated priors assign more mass to sparse matrices","Prior truncation affects sparsity in positive-definite inference"]},"model":"grok-4.3","cost_usd":0.00669,"raw_usage":{"total_tokens":3080,"prompt_tokens":593,"num_sources_used":0,"completion_tokens":55,"cost_in_usd_ticks":66899500,"prompt_tokens_details":{"text_tokens":593,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2432,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":593,"tokens_out":55,"duration_ms":22222,"temperature":1.0,"reasoning_tokens":2432,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-30T16:14:07.250343+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Compute or estimate the prior probability of matrices with a given number of zero off-diagonal entries under both the truncated and untruncated versions for fixed variance parameters as dimension increases; persistent deviation from equality for sparser cases would support the claim.","supporting_citations":[],"review_version":2}