{"id":"5e825aaf-2a8e-40dc-8136-c043c48ef076","arxiv_id":"2606.28652","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"AIHT is an online sparse quantile regression method with adaptive thresholding that achieves logarithmic regret under restricted curvature and gradient-leakage conditions.","lead":"The paper proposes Adaptive Iterative Hard Thresholding (AIHT), an online algorithm for high-dimensional quantile regression that alternates stochastic subgradient steps with adaptively timed hard-thresholding projections. A generalist might read it for methods that maintain sparsity while handling sequential data and nonsmooth losses like quantiles.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Validity of restricted curvature and gradient-leakage conditions for nonsmooth quantile loss","rationale":"The reader's weakest_assumption correctly isolates the key conditional step. The claim is explicitly 'under' those conditions, so the internal logic may be sound once the conditions are granted; the open question is their applicability, which the abstract does not resolve.","tokens_in":1649,"tokens_out":254,"duration_ms":22487,"concrete_test":"Generate a simple high-dimensional quantile regression instance (e.g., linear model with heavy-tailed noise) that matches the problem setup but violates the gradient-leakage condition; run AIHT and check whether it exits the sparse cone or loses the logarithmic regret bound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim states that under restricted curvature and gradient-leakage conditions, AIHT stays in an inflated sparse cone, shows two-phase convergence, and attains logarithmic regret. These conditions are invoked directly for all three properties but are not standard for quantile regression (nonsmooth pinball loss, possible heavy tails or heterogeneity). The abstract provides no discussion of when they hold or additional assumptions needed to ensure them, making this the least secure link in the argument.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript proposes Adaptive Iterative Hard Thresholding (AIHT) for online high-dimensional quantile regression. It alternates stochastic subgradient updates with adaptively scheduled hard-thresholding steps to separate support discovery from refinement. Under restricted curvature and gradient-leakage conditions, the algorithm is claimed to remain in an inflated sparse cone, exhibit two-phase convergence, and attain logarithmic regret for the sliding-window objective. Supporting simulations and threshold-scheduling ablations are provided.","tokens_in":1745,"tokens_out":378,"duration_ms":22106,"significance":"If the theoretical claims hold, this work would offer a principled approach to sparse online quantile regression in high dimensions, addressing challenges from nonsmooth loss and potential heterogeneity. The adaptive scheduling and two-phase behavior are novel aspects. The simulations provide empirical support for the mechanism.","major_comments":[{"comment":"Abstract: The central theoretical results (inflated sparse cone membership, two-phase convergence, and logarithmic regret) all rely on restricted curvature and gradient-leakage conditions, yet the abstract provides no discussion of when these hold for the nonsmooth pinball loss, nor any sufficient conditions on the data-generating process (e.g., for heavy tails or heterogeneity).","section":"Abstract"},{"comment":"Theory section: The regret bound is stated to follow directly from the curvature and leakage conditions; without an explicit argument showing these conditions are independently verifiable for quantile regression (rather than implicitly fitted to the target bound), the analysis risks circularity.","section":"Theory"}],"minor_comments":[{"comment":"The abstract refers to 'simulations for online quantile regression, together with threshold-scheduling ablations' but does not specify dimensions, sample sizes, or exact baselines, which would strengthen the empirical section.","section":null}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the detailed and constructive feedback. We address each major comment below and will revise the manuscript accordingly to improve clarity on the assumptions and their verifiability.","responses":[{"response":"We agree that the abstract would benefit from additional context on the assumptions. In the revision we will expand the abstract to note that the restricted curvature and gradient-leakage conditions are verifiable under standard assumptions on the design (restricted eigenvalue-type properties) and on the conditional distribution (bounded density in a neighborhood of the quantile), which accommodate heterogeneity and moderate heavy tails for the pinball loss.","revision_made":"yes","referee_comment":"[Abstract] Abstract: The central theoretical results (inflated sparse cone membership, two-phase convergence, and logarithmic regret) all rely on restricted curvature and gradient-leakage conditions, yet the abstract provides no discussion of when these hold for the nonsmooth pinball loss, nor any sufficient conditions on the data-generating process (e.g., for heavy tails or heterogeneity)."},{"response":"The conditions are defined independently of the regret bound and draw on standard restricted strong convexity notions for nonsmooth losses. To address the concern directly, the revised theory section will include a new remark or short subsection that states explicit sufficient conditions on the data-generating process (e.g., sub-exponential tails and positive density at the quantile) under which both curvature and leakage hold for the pinball loss, thereby separating the assumption verification from the regret derivation.","revision_made":"yes","referee_comment":"[Theory] Theory section: The regret bound is stated to follow directly from the curvature and leakage conditions; without an explicit argument showing these conditions are independently verifiable for quantile regression (rather than implicitly fitted to the target bound), the analysis risks circularity."}],"tokens_in":1272,"tokens_out":392,"duration_ms":14031,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The adaptive scheduling is the concrete new piece: delay thresholding early so weak signals can accumulate, then ramp up the frequency later to stabilize the sparse estimate. That mechanism is not the default in prior online IHT work and is backed by simulations that include threshold-scheduling ablations showing gains over standard baselines.\n\nThe theory claims that under restricted curvature and gradient-leakage conditions the iterates stay in an inflated sparse cone, show two-phase convergence, and achieve logarithmic regret on the sliding-window objective. The abstract states these results for the nonsmooth pinball loss, which is the right loss for quantile regression.\n\nThe soft spot is exactly those conditions. They are invoked for every major claim yet are not standard for quantile regression with possible heavy tails or heterogeneity, and the abstract gives no discussion of when they hold or what extra assumptions would make them plausible. Without that, the regret bound and cone-staying argument rest on an unverified link. The simulations are helpful for the scheduling idea but do not test the conditions themselves.\n\nThis is for people working on online sparse methods or streaming quantile regression. A reader who already knows IHT and wants a practical tweak for the online quantile case can extract the scheduling rule and the simulation comparisons. The paper is coherent on its own terms and engages the literature honestly, so it deserves a serious referee even though the assumption gap needs attention in revision.","headline":"The paper's main move is an adaptive schedule for hard-thresholding frequency in online IHT applied to high-dimensional quantile regression, separating early support discovery from later refinement.","tokens_in":2201,"tokens_out":357,"would_cite":false,"duration_ms":18299,"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":"Adaptive Iterative Hard Thresholding separates support discovery from refinement to achieve two-phase convergence and logarithmic regret in online high-dimensional quantile regression.","keywords":["online learning","high-dimensional regression","quantile regression","sparse estimation","iterative hard thresholding","regret bounds","stochastic subgradient"],"falsifier":"A dataset or simulation where the restricted curvature condition fails and the algorithm either leaves the inflated sparse cone or fails to achieve logarithmic regret on the sliding-window objective.","tokens_in":2547,"feed_emoji":"","tokens_out":650,"duration_ms":23415,"temperature":0.7,"pith_summary":"The paper proposes Adaptive Iterative Hard Thresholding as an online method for sparse quantile regression that alternates stochastic subgradient steps with hard-thresholding projections whose frequency increases over time. Early low-frequency thresholding gives weak but relevant coordinates time to build signal, while later high-frequency projections stabilize the sparse solution and exploit local curvature. Under restricted curvature and gradient-leakage conditions the iterates stay inside an inflated sparse cone, the algorithm shows an initial discovery phase followed by a refinement phase, and the sliding-window objective incurs only logarithmic regret. A reader would care because quantile loss is nonsmooth and standard online sparse methods either lose sparsity or suffer worse regret on heterogeneous or heavy-tailed streaming data.","feed_headline":"Adaptive hard thresholding attains logarithmic regret for online quantile regression","feed_subtitle":"Delaying then accelerating projection frequency keeps the estimator sparse while delivering two-phase convergence under curvature conditions","key_machinery":"Adaptive scheduling of hard-thresholding frequency that first delays then accelerates projection to separate support discovery from local refinement.","core_discovery":"AIHT alternates stochastic subgradient updates with adaptively scheduled hard-thresholding steps. By delaying thresholding early to accumulate signal in weak coordinates and increasing projection frequency later, the method maintains an inflated sparse cone, exhibits two-phase convergence, and attains logarithmic regret for the sliding-window objective under restricted curvature and gradient-leakage conditions.","pith_inferences":["The same early-delay-then-accelerate schedule could be tested on other online sparse problems such as logistic or hinge loss.","If the curvature condition holds only locally, a hybrid method that switches to full gradient steps after support stabilization might further reduce regret.","The two-phase behavior suggests that regret bounds for other iterative thresholding algorithms could be tightened by making the projection frequency data-dependent.","Real-time streaming applications with drifting distributions would require checking whether the sliding-window regret still controls performance on the most recent data."],"forward_implications":["The estimator remains inside an inflated sparse cone for the entire online process.","Convergence occurs in an early discovery phase followed by a later refinement phase.","Logarithmic regret holds for the sliding-window quantile objective.","The framework applies to nonsmooth losses with possible heterogeneity or heavy-tailed noise.","Ablation studies on threshold scheduling confirm the mechanism improves over fixed-frequency baselines."],"fun_headline_variants":["AIHT delays thresholding early for signal accumulation in online quantile regression","Two-phase convergence from adaptive hard thresholding in high-dimensional online regressio","AIHT attains logarithmic regret with scheduled projections for quantile regression","Adaptive thresholding stabilizes sparse cone under curvature in online quantile regression"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The data-generating process satisfies the restricted curvature and gradient-leakage conditions.","fun_headline_variants_meta":{"raw":{"variants":["AIHT delays thresholding early for signal accumulation in online quantile regression","Two-phase convergence from adaptive hard thresholding in high-dimensional online regression","AIHT attains logarithmic regret with scheduled projections for quantile regression","Adaptive thresholding stabilizes sparse cone under curvature in online quantile regression"]},"model":"grok-4.3","cost_usd":0.0038,"raw_usage":{"total_tokens":1931,"prompt_tokens":607,"num_sources_used":0,"completion_tokens":67,"cost_in_usd_ticks":37999500,"prompt_tokens_details":{"text_tokens":607,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1257,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":607,"tokens_out":67,"duration_ms":9337,"temperature":1.0,"reasoning_tokens":1257,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-30T09:48:45.499361+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A dataset or simulation where the restricted curvature condition fails and the algorithm either leaves the inflated sparse cone or fails to achieve logarithmic regret on the sliding-window objective.","supporting_citations":[],"review_version":1}