{"id":"581c24d7-cc89-440a-87e5-b37305f27c41","arxiv_id":"2606.20141","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"DASH reduces problem dimensionality in MIQP subset selection to improve MIP solver incumbent quality on hard portfolio instances.","lead":"The paper introduces DASH, a dimensionality reduction technique for large-scale convex mixed-integer quadratic programs applied to subset portfolio selection. It claims to deliver better incumbent solutions from MIP solvers like Gurobi on difficult instances defined by covariance condition number and weight constraints.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Difficulty metric (condition number + box constraints) lacks independent validation as predictor of Gurobi performance","rationale":"The reader already isolated the same load-bearing premise; the full-text placeholder does not alter the fact that the abstract (and therefore the claim) rests on an unvalidated difficulty proxy. No other internal inconsistency is visible from the supplied material.","tokens_in":1698,"tokens_out":284,"duration_ms":12049,"concrete_test":"Re-run the full experimental suite with DASH disabled; regress Gurobi time-to-incumbent and final gap against condition number (and box-constraint tightness) while controlling for problem size; if the R² is low or the slope insignificant, the premise that these quantities capture difficulty fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline claim requires that (i) condition number of the covariance and presence of box constraints are reliable, exogenous proxies for MIQP difficulty and (ii) the tested instances are representative of the target subclass. The abstract states difficulty “is related to” these quantities and that improvement “scales with” difficulty, but supplies no separate evidence that these quantities predict Gurobi run-time or integrality gap on the same instances when DASH is disabled. If the correlation is only observed inside the DASH runs, the scaling claim is circular and the representativeness assumption remains untested.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper proposes DASH (Decreasing Active Set Hierarchy), a dimensionality reduction method for large-scale convex MIQPs in subset portfolio selection. It claims that DASH improves the quality of incumbent solutions obtained by Gurobi, with the improvement being consistent, significant, and scaling with problem difficulty, where difficulty is associated with the condition number of the covariance matrix and box constraints on portfolio weights.","tokens_in":1797,"tokens_out":361,"duration_ms":31135,"significance":"If the method is shown to preserve solution quality and the experimental results hold under independent validation of the difficulty metric, this could offer a valuable approach for tackling computationally challenging MIQPs in finance. The focus on practical incumbent solutions for NP-hard problems is relevant.","major_comments":[{"comment":"The assertion that 'the magnitude and duration of improvement by DASH scale with the difficulty of the problem' is based on an untested assumption that condition number and box constraints are reliable predictors of Gurobi performance. The manuscript does not report separate experiments correlating these quantities with solver metrics in the absence of DASH, raising the risk that the scaling observation is circular.","section":"Abstract"},{"comment":"The description of DASH lacks sufficient detail on the dimensionality reduction procedure, including any pseudocode, mathematical formulation of how active sets are decreased, and proof or argument that optimality or feasibility is preserved for the original problem.","section":"Method"}],"minor_comments":[{"comment":"No quantitative results, specific metrics, or problem dimensions are provided despite claiming 'extensive set of numerical experiments'.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The paper's emphasis on computational experiments for MIQP may be better suited to an optimization or operations research journal rather than stat.CO."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for their constructive comments and recommendation for major revision. We address each major comment below and commit to revisions that strengthen the manuscript's clarity and rigor.","responses":[{"response":"We acknowledge the concern regarding potential circularity. While our experiments vary the condition number and box constraints and demonstrate corresponding scaling in DASH's benefits, we did not include separate runs isolating these metrics' correlation with baseline Gurobi performance. In the revision, we will add a dedicated subsection with experiments that directly correlate condition number and box constraint tightness with Gurobi metrics (e.g., time to reach target incumbent quality) in the absence of DASH, providing independent validation of the difficulty predictors.","revision_made":"yes","referee_comment":"[Abstract] The assertion that 'the magnitude and duration of improvement by DASH scale with the difficulty of the problem' is based on an untested assumption that condition number and box constraints are reliable predictors of Gurobi performance. The manuscript does not report separate experiments correlating these quantities with solver metrics in the absence of DASH, raising the risk that the scaling observation is circular."},{"response":"We agree that the method description requires expansion for reproducibility and rigor. The revised manuscript will include: (i) pseudocode for the full DASH procedure, (ii) explicit mathematical formulation of the active-set hierarchy and dimensionality reduction steps, and (iii) a formal argument establishing that the reduced formulation preserves feasibility of the original MIQP (solutions remain feasible when lifted back) while noting that DASH is intended as a heuristic to improve incumbent quality rather than to guarantee global optimality of the original problem.","revision_made":"yes","referee_comment":"[Method] The description of DASH lacks sufficient detail on the dimensionality reduction procedure, including any pseudocode, mathematical formulation of how active sets are decreased, and proof or argument that optimality or feasibility is preserved for the original problem."}],"tokens_in":1257,"tokens_out":415,"duration_ms":20357,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"DASH is a dimensionality reduction hierarchy for a subclass of convex MIQPs that arise in best-subset portfolio selection. The paper claims it improves the quality and speed of incumbent solutions from Gurobi when the problems are large and difficult, with the size of the gain scaling with problem difficulty.\n\nWhat stands out is the experimental design. The authors vary problem size, covariance condition number, and box constraints on weights, then show consistent improvements over plain Gurobi that grow as those difficulty indicators increase. That pattern is concrete and directly relevant to practitioners who already use commercial solvers on these models.\n\nThe soft spots are in the supporting material. The abstract contains no derivation of DASH, no pseudocode, and no argument that the reduction preserves optimality or feasibility. On the difficulty claim, the text ties improvement to condition number and box constraints but does not show separate evidence that these quantities predict Gurobi run time or gap when DASH is turned off. The stress-test note is accurate on this point from the abstract alone; without that check the scaling result could be circular. The full paper may address these gaps, but they are not visible here.\n\nThis paper is for people working on large-scale portfolio subset selection who need better practical performance from existing MIP solvers. A reader in operations research or quantitative finance would get value from the reported gains if the method turns out to be reproducible and the guarantees hold.\n\nIt deserves a serious referee. The application is narrow but real, and the empirical claims are specific enough to evaluate. I would send it to peer review rather than desk reject so the technical details and validation can be checked.","headline":"DASH is a targeted reduction for convex MIQPs in subset portfolio selection that reports better Gurobi incumbents on hard instances, but the abstract supplies no derivation, guarantees, or independent check on the difficulty metric.","tokens_in":2262,"tokens_out":419,"would_cite":false,"duration_ms":19931,"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":"DASH dimensionality reduction improves incumbent solutions from Gurobi on difficult large-scale MIQPs for subset portfolio selection.","keywords":["MIQP","dimensionality reduction","subset selection","portfolio optimization","mixed integer programming","incumbent solutions","Gurobi","convex optimization"],"falsifier":"An experiment comparing DASH and plain Gurobi on portfolio instances with low condition number but added difficulty from other sources, such as very large cardinality constraints, to check if the improvement disappears.","tokens_in":2586,"feed_emoji":"📉","tokens_out":559,"duration_ms":14409,"temperature":0.7,"pith_summary":"This paper introduces DASH, a method to reduce the dimensionality of large convex mixed-integer quadratic programs arising in best-subset selection. The approach is tested on portfolio optimization where the goal is to select a subset of assets under quadratic risk terms. When the covariance matrix is ill-conditioned or portfolio weights have box constraints, standard solvers like Gurobi struggle to find good solutions quickly. DASH consistently finds better incumbents in those cases, and the size of the gain grows with how hard the instance is.","feed_headline":"DASH cuts dimensions to lift Gurobi incumbents on tough MIQPs","feed_subtitle":"The reduction method delivers larger and longer-lasting gains precisely when covariance matrices are ill-conditioned and weights are boxed.","key_machinery":"DASH, or Decreasing Active Set Hierarchy, which hierarchically reduces the active set of variables in the MIQP to lower its dimensionality for faster solver progress.","core_discovery":"The central claim is that applying the Decreasing Active Set Hierarchy reduces the effective size of the MIQP while maintaining feasibility and objective quality, allowing MIP solvers to produce superior incumbent solutions for subset portfolio selection problems whose difficulty is governed by covariance condition number and weight bounds.","pith_inferences":["DASH might extend to other combinatorial subset problems beyond portfolios.","Preprocessing to estimate condition number could decide when to apply DASH.","Combining DASH with other cutting planes or heuristics could yield further gains."],"forward_implications":["Improvement is larger and lasts longer as problem difficulty increases.","DASH works best when the covariance matrix has a high condition number.","It provides practical value for large instances where global optimality cannot be reached.","The method targets convex MIQPs in subset selection."],"fun_headline_variants":["DASH reduces dimensions for Gurobi gains on hard MIQPs","DASH shrinks MIQP size to aid solver in portfolio selection","DASH helps Gurobi on MIQPs with ill conditioned covariances","DASH cuts effective MIQP size for stronger solver incumbents","DASH hierarchy reduces MIQP scale when weights are boxed"],"cache_read_input_tokens":64,"weakest_assumption_plain":"The assumption that problem difficulty for these MIQPs is reliably indicated by the condition number of the covariance matrix together with the box constraints on weights.","fun_headline_variants_meta":{"raw":{"variants":["DASH reduces dimensions for Gurobi gains on hard MIQPs","DASH shrinks MIQP size to aid solver in portfolio selection","DASH helps Gurobi on MIQPs with ill conditioned covariances","DASH cuts effective MIQP size for stronger solver incumbents","DASH hierarchy reduces MIQP scale when weights are boxed"]},"model":"grok-4.3","cost_usd":0.0106,"raw_usage":{"total_tokens":4653,"prompt_tokens":611,"num_sources_used":0,"completion_tokens":80,"cost_in_usd_ticks":105999500,"prompt_tokens_details":{"text_tokens":611,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3962,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":611,"tokens_out":80,"duration_ms":34326,"temperature":1.0,"reasoning_tokens":3962,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-26T14:59:31.910481+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"An experiment comparing DASH and plain Gurobi on portfolio instances with low condition number but added difficulty from other sources, such as very large cardinality constraints, to check if the improvement disappears.","supporting_citations":[],"review_version":1}