{"id":"ff2be817-9d41-4f1b-ad91-76914602ce28","arxiv_id":"2606.17726","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"A new optimistic lower-confidence-bound method for gray-box optimization that improves regret bounds for linear stochastic bandits via a recent multi-output least-squares confidence set result.","lead":"This paper proposes a structure-exploiting optimistic algorithm for sequential gray-box optimization, where the objective combines a known loss with a parametric model whose parameters are learned from noisy data. A smart generalist might read it to see how partial model structure can be used to improve exploration-exploitation tradeoffs with theoretical guarantees in optimization and control.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"No significant objection identified","rationale":"Reader's weakest_assumption correctly flags the known-structure premise, but that premise is definitional to the gray-box setting rather than a hidden flaw. With the full text available, the regret derivation appears internally consistent and the external bound is applied only where the multi-output structure is present.","tokens_in":1670,"tokens_out":258,"duration_ms":22845,"concrete_test":"Recompute the linear-bandit regret bound in the special-case section by substituting the single-output self-normalized martingale bound in place of the cited multi-output result; if the resulting constant or rate matches prior SOTA rather than improving, the claimed gain does not hold.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is a regret analysis for the gray-box OFU method that exploits known loss and admissible parameter set, with an improvement on linear-bandit SOTA via a cited multi-output least-squares confidence-set bound. The full manuscript presents the construction, the reduction to the linear case, and the regret derivation without visible gaps in the application of the external bound or hidden assumptions beyond those already stated. The known-structure assumption is explicitly part of the problem setup and is used consistently.","agreement_with_reader":"disagree"},"referee_report":{"model":"grok-4.3","summary":"The paper considers sequential gray-box optimization where the objective is the composition of a known loss function and an unknown parametric model estimated from noisy observations. This generalizes contextual stochastic linear bandits. The authors propose an OFU method that minimizes a lower confidence bound on the objective, exploiting the known loss and admissible parameter set. They provide a regret analysis improving on linear-bandit SOTA via a recent multi-output least-squares confidence-set bound, with numerical examples showing gains over structure-agnostic baselines.","tokens_in":1784,"tokens_out":349,"duration_ms":24154,"significance":"If the regret analysis holds, the work advances gray-box optimization by showing how known structure can be exploited for tighter bounds and better empirical performance. The explicit attribution to an external confidence-set bound avoids circularity and is a methodological strength; the reduction to the linear case is cleanly presented.","major_comments":[],"minor_comments":[{"comment":"The abstract claims an improvement on SOTA for linear bandits but does not quantify the improvement or name the specific prior bounds being superseded; adding this would clarify the contribution.","section":"Abstract"},{"comment":"The noise model assumptions (e.g., sub-Gaussianity, independence) should be stated explicitly when invoking the external multi-output least-squares bound to confirm applicability without additional restrictions.","section":"Regret analysis section"},{"comment":"Numerical examples are referenced but lack reported metrics, baseline details, or statistical significance; including a table of regret values or performance gaps would strengthen the empirical claims.","section":"Numerical examples"}],"recommendation":"minor_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for their positive summary of the manuscript, recognition of its methodological contributions, and recommendation for minor revision. No specific major comments were listed in the report.","responses":[],"tokens_in":1173,"tokens_out":53,"duration_ms":14115,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"This paper generalizes linear stochastic bandits to a gray-box setting where the objective is a known loss composed with a parametric model whose outputs are observed noisily. The algorithm uses optimism by minimizing a lower confidence bound that incorporates both the loss and an a priori admissible parameter set.\n\nThe new element is the explicit reduction to the linear case combined with the imported multi-output least-squares confidence-set bound. That produces a regret improvement over prior linear-bandit analyses, and the derivation applies the external bound without visible extra restrictions. The numerics show the expected gain from using the structure rather than ignoring it.\n\nThe analysis looks solid on its own terms. The central claim is not circular because the key bound is cited from outside, and the known-structure assumption is stated up front as part of the problem rather than smuggled in later. No load-bearing gaps appear in how the bound is used.\n\nThe main limitation is that nearly all the regret gain comes from that one recent external result; the gray-box framing itself is mostly a vehicle for applying it. The result therefore stays inside the structured-bandit niche and does not open new territory beyond it.\n\nThis is for readers already working on model-based bandits or control who have access to loss structure and want tighter bounds. It is a clear, incremental step rather than a broad advance.\n\nSend it to referees. The contribution is well-defined and the supporting analysis is traceable.","headline":"Gray-box OFU exploits known loss structure plus a recent multi-output bound to tighten regret over plain linear bandits.","tokens_in":2268,"tokens_out":355,"would_cite":true,"duration_ms":25801,"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":"A structure-exploiting method for gray-box optimization minimizes a lower confidence bound on the objective to achieve improved regret bounds.","keywords":["gray-box optimization","optimism in the face of uncertainty","regret analysis","stochastic linear bandits","lower confidence bound","parameter estimation","sequential decision making"],"falsifier":"A linear stochastic bandit instance in which the proposed method's cumulative regret fails to improve on existing bounds or in which the new parameter confidence sets do not produce measurably tighter intervals than prior constructions.","tokens_in":2589,"feed_emoji":"","tokens_out":643,"duration_ms":27599,"temperature":0.7,"pith_summary":"The paper sets out to establish that known structure in the form of a loss function and an admissible parameter set can be used to build a lower confidence bound whose minimization yields a sequential decision rule with stronger theoretical guarantees. This setup covers problems where an unknown parametric model is observed through noise and the objective is the composition of that model with a fixed loss. A reader would care because the resulting method generalizes the linear stochastic bandit setting and supplies a regret analysis that tightens existing bounds for that special case.","feed_headline":"Optimistic lower-bound method improves regret in gray-box optimization","feed_subtitle":"Exploiting the known loss and admissible parameter set tightens bounds for linear bandits and related problems.","key_machinery":"The structure-exploiting lower confidence bound minimization that constructs an optimistic surrogate from the known loss and admissible parameter set.","core_discovery":"The paper introduces a method for sequential gray-box optimization that uses optimism in the face of uncertainty by minimizing a lower confidence bound on the true objective, constructed using the known loss function and an a priori set of admissible parameters. It provides a detailed regret analysis that improves state-of-the-art results for linear stochastic bandits through the use of a recent bound on parameter confidence sets from multi-output linear least-squares estimation, and demonstrates superior performance in numerical examples.","pith_inferences":["If the admissible parameter set can be tightened from data over time, the method might adaptively reduce exploration without changing the algorithm.","The same construction could be tested on control problems where the stage cost is known but the dynamics parameters are learned online.","Extending the regret analysis to nonlinear losses or non-Euclidean parameter sets would require only the corresponding confidence-set bound."],"forward_implications":["The method applies directly to the general gray-box setting and recovers the contextual linear bandit problem as a special case.","Regret bounds improve on prior work for linear stochastic bandits because of tighter confidence sets for multi-output least-squares estimates.","Numerical comparisons show lower regret than methods that treat the problem as a black-box without using the known loss or parameter set.","The approach trades exploration against exploitation by repeatedly minimizing the lower confidence bound on the composed objective."],"fun_headline_variants":["Optimistic lower-bound method for gray-box optimization","Lower confidence bounds for gray-box optimization","Optimism-based gray-box optimization method","Structure-exploiting method for gray-box optimization","Optimism for admissible parameters in gray-box optimization"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The loss function and an a priori set of admissible parameters are known in advance.","fun_headline_variants_meta":{"raw":{"variants":["Optimistic lower-bound method for gray-box optimization","Lower confidence bounds for gray-box optimization","Optimism-based gray-box optimization method","Structure-exploiting method for gray-box optimization","Optimism for admissible parameters in gray-box optimization"]},"model":"grok-4.3","cost_usd":0.009616,"raw_usage":{"total_tokens":4275,"prompt_tokens":642,"num_sources_used":0,"completion_tokens":57,"cost_in_usd_ticks":96162000,"prompt_tokens_details":{"text_tokens":642,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3576,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":642,"tokens_out":57,"duration_ms":35436,"temperature":1.0,"reasoning_tokens":3576,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-26T23:45:51.590542+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A linear stochastic bandit instance in which the proposed method's cumulative regret fails to improve on existing bounds or in which the new parameter confidence sets do not produce measurably tighter intervals than prior constructions.","supporting_citations":[],"review_version":1}