{"id":"b4ec4d7b-65c9-43da-85fe-45fe2e839b2b","arxiv_id":"2508.16588","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Allowing an adversarial-RL market maker to occasionally refuse quoting or quote single-sided improves returns and Sharpe ratios while keeping quoting-ratio compliance near 99.9%.","lead":"This paper studies market-making algorithms trained with adversarial reinforcement learning and asks whether letting the agent occasionally quote nothing, or only one side of the market, beats always quoting both sides. In simulations, allowing the trader to choose when to quote improves returns and Sharpe ratios while staying above market quoting requirements.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Load-bearing risk: the simulated advantage of refusing to quote may be an artifact of omitting real absence costs (queue priority, re-entry latency, information leakage); the provided full text is corrupted/unusable, so this cannot be checked.","rationale":"I read the abstract's claim as a comparative statement: in adversarial simulated markets, adding no-quote/single-sided actions to an RL MM improves returns/Sharpe while satisfying quoting ratios. For this to be true and useful, the simulation must capture the costs a real MM faces when it stops quoting. The abstract does not mention such costs, and the only full-text block provided is corrupted and carries a different arXiv ID, so there is no way to check whether they are in the simulator. The concern is not that the in-simulation result is internally false; with a larger action space, no-quote agents should weakly dominate the continuous baseline, and the abstract's 'and/or' hints the improvement may be metric-selective. The open question is whether the improvement transfers. A re-entry-cost ablation would settle this. I therefore keep the reader's UNVERDICTED verdict unchanged rather than moving to reject: the concern points to a missing verification, not a demonstrated failure.","tokens_in":6307,"tokens_out":6215,"duration_ms":68244,"concrete_test":"Download the actual source/PDF of arXiv:2508.16588, locate the environment definition, and check whether the state/transition includes queue position or a re-entry cost. Then run the same three agents in an otherwise identical environment with a single modification: any transition from 'no quote' or 'single-sided' back to two-sided quoting resets the MM's queue priority to the back (or incurs a fixed cost proportional to the spread). Recompute the returns/Sharpe comparison. If the selective-quoting advantage disappears or reverses, the original result is an artifact of free re-entry; if it survives, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim needs the environment to make 'not quoting' costless apart from a quoting-ratio requirement. The abstract explicitly frames the only obligation as market-rule quoting ratios, and the evaluation is in adversarial model-driven environments. If the simulator's state/transition model does not penalize an agent that stops quoting and later returns, the no-quote action lets the RL agent skip high-adverse-selection ticks and resume with no loss of queue position or latency. The reported improvement in returns/Sharpe then follows mechanically from expanding the action space and does not establish a robust market-making strategy. No available text describes queue priority, re-entry costs, or information effects of absence. The attached 'full text' is corrupted mojibake and even carries a different arXiv identifier (2508.16586, physics.bio-ph), so the methods/reward/parameters cannot be inspected. This is the key correctness risk: if the missing cost is added, the advantage may vanish; if not, the claim is simulator-specific.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims that enriching the action space of an adversarial-RL-trained market maker with the option to occasionally refrain from quoting, or to quote only one side, improves returns and/or Sharpe ratios relative to a continuously quoting market maker, while still satisfying market-rule quoting ratios. Two new agents are introduced: one that can either quote both sides or refuse to quote, and one that can also quote only the bid or only the ask. The comparison is carried out in model-driven adversarial environments built on the authors' prior work. The abstract reports improvements and quoting ratios reaching up to 99.9% in some cases.","tokens_in":6410,"tokens_out":2543,"duration_ms":29983,"significance":"If the central claim is correct, the paper identifies a practically relevant and underexplored degree of freedom in automated market making: the ability to skip quoting in adverse conditions without violating regulatory quoting obligations. The extension of the action space is a natural and potentially valuable contribution to the adversarial-RL market-making literature, and the two new agent designs are clearly described in the abstract. However, the significance cannot currently be assessed because the technical content needed to validate the claim is not available in the supplied manuscript: the body text is corrupted, no methods, environment equations, reward definitions, or experimental protocols are readable, and no statistical evidence is reported. The contribution is therefore plausible but unverified.","major_comments":[{"comment":"The supplied full text is largely unreadable mojibake and even carries a different arXiv identifier (2508.16586, physics.bio-ph) rather than the paper's stated identifier. I cannot locate the environment dynamics, state/transition model, reward function, action parameterization, training hyperparameters, or evaluation protocol. This is a load-bearing gap: the central claim that refusing to quote improves returns depends on whether the simulator imposes costs on absence such as queue priority loss, re-entry latency, or information leakage. As it stands, the advantage of the no-quote action may be mechanical if the only obligation is a quoting-ratio constraint with no other absence costs. The authors must supply a clean, readable manuscript with full methods before the claim can be evaluated.","section":"Full text / all sections"},{"comment":"The abstract's empirical claims are hedged and lack quantitative support: 'improves returns and/or Sharpe ratios' is not accompanied by any effect size, number of seeds, confidence intervals, significance tests, or environment parameter values. The phrase 'reaching up to 99.9% in some cases' is a best-case selection and does not indicate typical or worst-case quoting ratios. The comparison baseline ('continuously quoting MM') is not defined in terms of reward specification or training procedure. These omissions prevent the reader from assessing robustness and are central to the paper's contribution.","section":"Abstract"},{"comment":"The evaluation appears to be performed in the same model-driven adversarial environments used for training, and those environments inherit from the authors' prior adversarial-RL line. This creates a risk of in-sample overfitting to the adversarial generator. The manuscript should report out-of-sample or cross-environment robustness checks, including sensitivity to reward weights for quoting-ratio compliance, disturbance distributions, and parameter ranges. Without such evidence, the claimed improvement may be an artifact of the specific simulator rather than a general property of allowing intermittent quoting.","section":"Evaluation / environments"}],"minor_comments":[{"comment":"The term 'quoting ratio' should be defined precisely: is it time-weighted, volume-weighted, or trade-weighted, and over what time window? Also specify the market rules that determine the required ratio.","section":"Abstract / quoting ratio"},{"comment":"'Basically meet any market requirements' is vague. The paper should report the achieved quoting ratios for all agents, across environments, rather than only the maximum case.","section":"Abstract"},{"comment":"The repeated formatting errors and the mismatched arXiv identifier in the supplied text need to be fixed before resubmission; the current version cannot be processed by a reader.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The supplied file appears to be a corrupted extraction with a mismatched arXiv identifier. The editor may wish to verify that the correct manuscript was submitted. My recommendation of major_revision reflects the absence of readable methods rather than any judgment about the underlying research idea; a clean, complete manuscript is a prerequisite for substantive review."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The one thing you should know: this paper's contribution is the action-space extension. Prior adversarial-RL market-making work assumes continuous quoting; here they add refuse-to-quote and single-sided quoting actions, and report improved returns/Sharpe. That is a real, well-scoped extension, and the practical hook (registered MMs only need to satisfy quoting ratios) is legitimate. I would give the authors credit for the idea even from the abstract alone.\n\nThe problem: the only thing I can evaluate is the abstract. The attached full text is corrupted mojibake and even carries a different arXiv ID (2508.16586, physics.bio-ph). So there are no methods, no environment parameters, no reward function, no number of seeds, no significance tests, no baseline definitions. The abstract itself hedges with 'improves returns and/or Sharpe ratios' and 'up to 99.9% in some cases.' That is best-case framing.\n\nThe soft spot that would determine correctness: the advantage of refusing to quote may be an artifact of the simulator not charging for absence. The abstract frames the only obligation as quoting ratios, and evaluation is in the same model-driven adversarial environments used for training. If the simulator lets an MM skip adverse-selection ticks and resume with no loss of queue priority, re-entry latency, or information leakage, the result follows mechanically from expanding the action space. That is not necessarily wrong—it could still be a useful robustness result for market-rule-constrained MMs—but it would be simulator-specific. I cannot check it because the text is unreadable.\n\nThe reader's take and stress-test note both hit this point, and I agree with them. The central claim is internally coherent and consistent with inventory-risk intuition, so this is not a pile of nonsense. But it is unverified as presented.\n\nWho is this for? Researchers in RL market making, and practitioners who want a principled reason to build selective-quoting policies. It is a moderate extension of the authors' prior adversarial-RL line, not a paradigm shift. If the body holds up, it deserves a serious referee. I would not desk-reject it solely on the abstract, because the idea is novel relative to the continuous-quoting assumption and the quoting-ratio constraint is a real-world obligation.\n\nGiven that the full text is corrupted and likely a pipeline artifact, my recommendation is: ask the authors to resubmit a readable PDF, then send to peer review. The paper deserves referee time.","headline":"A plausible extension of adversarial-RL market making with real practical appeal, but the provided full text is corrupted and the core claim hinges on whether the simulator charges for absence; worth a referee once the manuscript is readable.","tokens_in":7022,"tokens_out":2344,"would_cite":false,"duration_ms":21912,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A market maker trained to sometimes refuse quotes, or quote only one side, earns better returns and Sharpe ratios than one that quotes continuously, while still meeting quoting-ratio requirements.","keywords":["market making","adversarial reinforcement learning","quoting ratio","single-sided quotes","bid-ask spread","Sharpe ratio","action space","adversarial environments"],"falsifier":"Evaluate the trained refusal-enabled policy and the continuous-quoting baseline on a held-out set of adversarial environments never seen during training, or on a real order-book replay, with both agents forced to the same realized quoting ratio, such as 95%. If the refusal-enabled policy does not achieve better returns or Sharpe ratio than the baseline, the paper's central claim is not supported.","tokens_in":6064,"feed_emoji":"📈","tokens_out":5945,"duration_ms":57957,"temperature":0.7,"pith_summary":"Market making is usually modeled as a game of continuously posting both a buy and a sell quote. This paper relaxes that assumption: it gives a reinforcement-learning market maker the extra options to refuse to quote, or to post only one side, and tests the result in adversarial simulated markets. Its central claim is that these richer action spaces improve returns and/or Sharpe ratios relative to a continuously quoting market maker, while the agent still learns to satisfy any required quoting ratio, up to 99.9% in some cases. If true, this matters because real market makers, even registered ones, are not obliged to quote at every moment, and selective silence is a practical robustness lever that the previous modeling line had excluded.","feed_headline":"Sometimes refusing to quote beats always quoting","feed_subtitle":"Market makers trained to stay quiet or quote one side lift returns and Sharpe while still meeting quoting-ratio rules.","key_machinery":"The load-bearing mechanism is the expanded action space. Instead of forcing the market maker to always post both sides, each time step is a choice among two or four quoting modes, including silence and single-sided quotes; adversarial reinforcement learning turns that expanded space into a policy that withholds liquidity selectively. Quoting-ratio requirements enter through the environment or reward design, so the agent must balance the benefit of silence against the obligation to quote often. The continuously-quoting agent serves as the control that isolates the effect of the refusal option.","core_discovery":"The paper's central claim is that an automated market maker does not have to quote continuously to perform well under adverse conditions. It builds two new agents on top of a continuously-quoting baseline: one whose action set is {post bid and ask, refuse to quote}, and one whose action set is {post bid and ask, post only ask, post only bid, refuse to quote}. Both are trained with adversarial reinforcement learning in a model-driven simulator and compared with the baseline across several adversarial environments. The reported result is that occasional refusal improves returns and/or Sharpe ratios, and that learned quoting ratios are controllable and can reach 99.9% in some cases. The contrib","pith_inferences":["If the reported effect transfers to real markets, minimum-quoting regulations should be treated as an endogenous constraint on action frequency rather than a reason to force continuous quoting; a concrete next step is adding an explicit compliance term to the reward and measuring the trade-off.","The policy's choice of when to refuse is likely correlated with adverse-selection pressure; inspecting which simulated states trigger refusal—order-flow imbalance, spread width, recent losses—would make the mechanism testable rather than black-box.","Single-sided quoting may be doing inventory-management work, such as quoting only the side that reduces a skewed book; comparing inventory-dependent action frequencies with an inventory-targeting baseline would test this.","Because performance is measured inside the training simulator, the strongest extension would be evaluation on held-out adversarial environments or real order-book replays, which lies outside the paper's current scope."],"forward_implications":["If the claim holds, automated market makers can be made more resilient without changing inventories, pricing models, or exchange rules—only by allowing the policy to say no.","Quoting-ratio obligations become a tuneable constraint: the same agent can be steered toward high quoting rates, such as 99.9%, or lower ones, while retaining some or all of the performance gain.","Single-sided quoting is a meaningful intermediate tool, letting the agent stay present in the market while avoiding the more costly side of the book.","In the tested adversarial environments, the refusal-capable agents outperform the always-quoting baseline, isolating the richer action space as the source of the improvement."],"supporting_citations":[],"fun_headline_variants":["Quiet market makers beat constant quoters","Refusing to quote can boost trading returns","Selective quoting lifts Sharpe, meets rules","Market makers profit from occasional silence","To quote or not to quote: a winning strategy"],"cache_read_input_tokens":2816,"weakest_assumption_plain":"The load-bearing premise is that the simulated adversarial environments used for training and evaluation capture the real costs, volatility regimes, and order-flow adversities a market maker would face when deciding to stop quoting; if they do not, the measured gains may not transfer out of the simulator.","fun_headline_variants_meta":{"raw":{"variants":["Quiet market makers beat constant quoters","Refusing to quote can boost trading returns","Selective quoting lifts Sharpe, meets rules","Market makers profit from occasional silence","To quote or not to quote: a winning strategy"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000198,"raw_usage":{"total_tokens":1217,"prompt_tokens":769,"completion_tokens":448,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":513,"completion_tokens_details":{"reasoning_tokens":382}},"tokens_in":513,"tokens_out":448,"duration_ms":5215,"temperature":1.0,"reasoning_tokens":382,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T23:06:27.214728+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the trained refusal-enabled policy and the continuous-quoting baseline on a held-out set of adversarial environments never seen during training, or on a real order-book replay, with both agents forced to the same realized quoting ratio, such as 95%. If the refusal-enabled policy does not achieve better returns or Sharpe ratio than the baseline, the paper's central claim is not supported.","supporting_citations":[],"review_version":1}