{"id":"cde55252-0d7f-4a04-861b-f186b8b762fb","arxiv_id":"2605.24646","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Robust ergodic singular control for jump-diffusions with drift and intensity uncertainty reduces the HJB to a nonlinear integro-differential free-boundary problem whose worst-case model is bang-bang and whose optimal policy uses reflecting barriers; exponential jumps further reduce it to ODEs.","lead":"This paper formulates a robust ergodic singular control problem for jump-diffusion processes with uncertainty in both drift and jump intensity, using entropy-penalized distortions to handle model ambiguity. A smart generalist might read it for methods to design robust long-run policies in applications like inventory or cash management under incomplete information about shocks.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Well-posedness of the max-min ergodic problem (existence/uniqueness of value function satisfying the claimed HJB) is assumed rather than derived from first principles.","rationale":"The reader's weakest_assumption isolates precisely the step that must be true for every subsequent reduction (HJB form, bang-bang structure, ODE collapse) to be valid; without an explicit existence argument or verification theorem in the full text, this remains the load-bearing gap. No other internal inconsistency is visible from the abstract and claim structure.","tokens_in":1692,"tokens_out":350,"duration_ms":16511,"concrete_test":"Starting from the dynamic programming principle for the ergodic cost, derive the HJB variational inequality directly (without assuming the free-boundary ansatz) and check whether any C^2 solution to the reduced ODE system (under exponential jumps) satisfies the original integro-differential equation pointwise outside the continuation region; if the integro term fails to match after substituting the candidate worst-case intensity, the reduction does not hold.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest claim requires that the entropy-penalized max-min problem admits a value function whose HJB equation reduces exactly to the stated nonlinear integro-differential free-boundary problem with bang-bang worst-case controls and reflecting barriers. This hinges on the long-run average criterion being well-posed under simultaneous drift and intensity distortions; standard verification theorems for ergodic singular control do not automatically extend to jump-diffusions with relative-entropy penalties on the intensity measure, where the effective generator may lose ellipticity or the average cost may fail to be finite without additional growth or moment conditions on the jump kernel.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper formulates a robust ergodic singular control problem for jump-diffusion processes subject to uncertainty in both drift and jump intensity, using entropy-penalized distortions in a max-min long-run average criterion. It claims that the associated HJB equation reduces to a nonlinear integro-differential free-boundary problem whose worst-case controls are bang-bang and whose optimal policy is characterized by reflecting barriers; under exponentially distributed jumps the problem further reduces to a system of ODEs.","tokens_in":1822,"tokens_out":381,"duration_ms":22434,"significance":"If the claimed reductions and the underlying well-posedness hold, the work would supply a concrete, numerically tractable framework for robust singular control of jump-diffusions under model ambiguity, directly applicable to inventory, cash-management, and capacity-planning problems. The explicit reduction to ODEs under exponential jumps is a concrete computational advantage.","major_comments":[{"comment":"The central claim that the HJB equation reduces to the stated nonlinear integro-differential free-boundary problem with bang-bang worst-case controls and reflecting barriers presupposes existence and uniqueness of a value function for the max-min ergodic problem. No verification theorem, existence proof, or growth/moment conditions ensuring the average cost remains finite under simultaneous drift and intensity distortions are supplied; standard ergodic singular-control verification results do not automatically extend to relative-entropy penalties on the intensity measure.","section":"Abstract and the HJB reduction claim"},{"comment":"The reduction to a system of ODEs under exponentially distributed jumps is asserted without an explicit derivation or error analysis showing that the integro-differential terms collapse exactly to the claimed ODE system while preserving the free-boundary structure.","section":"Abstract (final sentence)"}],"minor_comments":[],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the careful review and constructive comments on our manuscript. The two major comments identify areas where additional rigor and explicit derivations would strengthen the presentation. We respond to each below and commit to revisions that address the concerns without altering the core claims.","responses":[{"response":"We agree that a self-contained verification argument is needed. The submitted manuscript derives the HJB reduction formally under the assumption that a sufficiently regular value function exists, but does not supply the supporting existence/uniqueness result or the moment conditions that guarantee finiteness of the ergodic cost under joint drift-intensity distortions. In the revision we will insert a new subsection that states explicit growth and integrability conditions on the jump measure (ensuring the relative-entropy penalty remains well-defined) and sketches a verification theorem that adapts standard ergodic singular-control arguments to the entropy-penalized intensity control; the argument will be referenced to related robust-control literature where appropriate.","revision_made":"yes","referee_comment":"[Abstract and the HJB reduction claim] The central claim that the HJB equation reduces to the stated nonlinear integro-differential free-boundary problem with bang-bang worst-case controls and reflecting barriers presupposes existence and uniqueness of a value function for the max-min ergodic problem. No verification theorem, existence proof, or growth/moment conditions ensuring the average cost remains finite under simultaneous drift and intensity distortions are supplied; standard ergodic singular-control verification results do not automatically extend to relative-entropy penalties on the intensity measure."},{"response":"The reduction is carried out in Section 4 by direct substitution of the exponential density into the integro-differential operator, followed by integration by parts that converts the nonlocal terms into local coefficients while leaving the free-boundary conditions unchanged. Because the substitution is exact, no approximation error arises. To make the algebra fully transparent we will add an appendix that reproduces the substitution step by step, verifies that the free-boundary structure is preserved, and confirms that the resulting system is indeed a set of ODEs with the same boundary conditions.","revision_made":"yes","referee_comment":"[Abstract (final sentence)] The reduction to a system of ODEs under exponentially distributed jumps is asserted without an explicit derivation or error analysis showing that the integro-differential terms collapse exactly to the claimed ODE system while preserving the free-boundary structure."}],"tokens_in":1325,"tokens_out":506,"duration_ms":25829,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper extends entropy-penalized robust control to an ergodic singular control problem on jump-diffusions, where both drift and jump intensity are uncertain. It claims the max-min problem yields an HJB that reduces to a nonlinear integro-differential free-boundary equation, with bang-bang worst-case controls and optimal policy given by reflecting barriers. For exponential jumps the whole thing collapses to a system of ODEs that can be solved numerically.\n\nThis combination is not routine. Handling simultaneous uncertainty on drift and intensity inside an ergodic criterion with singular interventions, then getting an explicit tractable structure, is the actual new piece. The application examples in inventory and cash management are standard but fit the setting cleanly.\n\nThe soft spot is exactly what the stress-test flags: the abstract states the reductions without derivations, verification theorems, or checks on whether the entropy penalty on intensity keeps the long-run average finite and the generator well-behaved. Standard ergodic singular control results do not automatically survive the intensity distortion, and nothing here shows the extra conditions that would be needed. If the full paper only assumes the HJB form rather than deriving it from first principles, the central claim stays unverified.\n\nThe work is aimed at people already doing robust stochastic control in operations research. A reader who wants general theory on jump-diffusions will not find it, but someone who needs a concrete robust policy for a jump model with numerics might get usable output.\n\nSend it to peer review. The formulation is reasonable and the claimed reductions look worth checking in detail; the subfield can decide if the technical gaps are filled once the derivations are on the table.","headline":"The paper reduces robust ergodic singular control for jump-diffusions with drift and intensity uncertainty to a free-boundary problem and ODEs for exponential jumps, but the abstract leaves the well-posedness steps unshown.","tokens_in":2324,"tokens_out":422,"would_cite":false,"duration_ms":32591,"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":"The Hamilton-Jacobi-Bellman equation for robust ergodic control of jump-diffusions reduces to a nonlinear integro-differential free-boundary problem.","keywords":["robust control","jump diffusion","ergodic control","singular control","free boundary problem","entropy penalty","model uncertainty","reflecting barriers"],"falsifier":"A direct verification that the candidate solution of the free-boundary problem fails to satisfy the original Hamilton-Jacobi-Bellman equation when the adversary chooses non-bang-bang distortions.","tokens_in":2584,"feed_emoji":"","tokens_out":644,"duration_ms":33820,"temperature":0.7,"pith_summary":"This paper addresses regulation of systems with both continuous noise and sudden jumps when the model parameters themselves are uncertain. It poses the problem as a robust ergodic singular control task in which a controller chooses interventions while an adversary distorts the drift and jump rates under an entropy penalty. The central result is that the associated Hamilton-Jacobi-Bellman equation collapses to a nonlinear integro-differential free-boundary problem whose solution determines the worst-case model and the optimal reflecting barriers. When the jumps follow an exponential distribution the free-boundary problem further simplifies to a finite system of ordinary differential equations that can be solved numerically. The approach therefore supplies a concrete computational route for finding robust long-run policies in settings such as inventory and cash management.","feed_headline":"Robust ergodic control reduces to free-boundary problem","feed_subtitle":"Entropy-penalized uncertainty in drift and jump intensity produces bang-bang worst-case models and reflecting barrier policies.","key_machinery":"The nonlinear integro-differential free-boundary problem that arises from the Hamilton-Jacobi-Bellman equation of the robust ergodic criterion.","core_discovery":"The authors establish that the max-min ergodic control problem for a jump-diffusion with uncertain drift and intensity measure reduces to a nonlinear integro-differential free-boundary problem. The worst-case distortions take a bang-bang form, the optimal policy consists of reflecting barriers, and the exponential-jump case yields an ordinary differential equation system.","pith_inferences":["The reflecting-barrier structure may persist for other jump distributions that preserve the tractability of the free-boundary problem.","Numerical solutions of the resulting ODEs could be used to quantify the effect of increasing model ambiguity on the width of the no-intervention region.","Similar entropy-penalized formulations might be applied to other ergodic control problems with jump components."],"forward_implications":["The worst-case model takes a bang-bang form.","The optimal policy is given by reflecting barriers.","Exponential jump distributions reduce the problem to a system of ordinary differential equations.","Applications include inventory control, cash management, and capacity planning."],"fun_headline_variants":["Jump-diffusion control with uncertainty reduces to free-boundary HJB","Ergodic singular control reduces to integro-differential free-boundary","Worst-case models show bang-bang form in jump-diffusion ergodic control","Reflecting barriers characterize optimal policies for uncertain jumps"],"cache_read_input_tokens":64,"weakest_assumption_plain":"Model ambiguity is captured exactly by entropy-penalized distortions of the drift and intensity measures and the long-run average criterion admits a well-posed max-min formulation.","fun_headline_variants_meta":{"raw":{"variants":["Jump-diffusion control with uncertainty reduces to free-boundary HJB","Ergodic singular control reduces to integro-differential free-boundary","Worst-case models show bang-bang form in jump-diffusion ergodic control","Reflecting barriers characterize optimal policies for uncertain jumps"]},"model":"grok-4.3","cost_usd":0.007304,"raw_usage":{"total_tokens":3324,"prompt_tokens":590,"num_sources_used":0,"completion_tokens":70,"cost_in_usd_ticks":73037000,"prompt_tokens_details":{"text_tokens":590,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2664,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":590,"tokens_out":70,"duration_ms":24900,"temperature":1.0,"reasoning_tokens":2664,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-30T13:03:28.375457+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A direct verification that the candidate solution of the free-boundary problem fails to satisfy the original Hamilton-Jacobi-Bellman equation when the adversary chooses non-bang-bang distortions.","supporting_citations":[],"review_version":1}