{"id":"2bcc9aee-5240-4f95-bf34-0f1e0a64ded4","arxiv_id":"2605.27658","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":0.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"This review summarizes the evolution of probability measures in asset pricing from early state price theory to contemporary learned pricing kernels.","lead":"This paper reviews the historical development of probability measures in asset pricing, from state prices and martingale measures to modern data-driven pricing kernels and information-adjusted forecasting. A smart generalist might read it to understand why asset pricing relies on constructing or selecting special probabilities rather than estimating real-world ones.","discovery_kind":"review","skeptic_critique":{"model":"grok-4.3","headline":"No significant objection identified","rationale":"Reader correctly flags the paper as exposition rather than research and notes the selection-bias risk in any historical narrative. Because the theme is uncontroversial and no new claims are made, that risk does not constitute a load-bearing concern for the central claim. Verdict therefore remains UNVERDICTED with no adjustment.","tokens_in":1854,"tokens_out":273,"duration_ms":11380,"concrete_test":"Cross-check one representative formula collection (e.g., the Radon-Nikodym / Girsanov / SDF identities listed in the abstract) against the corresponding primary references cited in the manuscript; if the formulas are reproduced without distortion, the review's technical content is accurate.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The manuscript is an expository historical review whose central theme—that asset pricing routinely constructs, transforms, or selects probability measures rather than estimating physical probabilities—is a standard organizing principle already present in textbooks (Duffie, Cochrane) and the cited landmark papers. No novel theorem, empirical claim, or derivation is advanced that could be internally inconsistent or rest on a fragile technical assumption. The listed milestones (Bachelier through modern ML kernels) are canonical; any selection bias is inherent to review format but does not undermine the descriptive claim.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript is a historical review of probability measures in asset pricing. It traces developments from Bachelier's probabilistic speculation models and Arrow-Debreu state-contingent claims, through Black-Scholes-Merton risk-neutral valuation, Harrison-Kreps/Harrison-Pliska martingale measures, Delbaen-Schachermayer fundamental theorems, Breeden-Litzenberger implied densities, change-of-numeraire techniques, Hansen-Jagannathan SDF bounds, Cochrane's SDF synthesis, and recent ML/data-driven pricing kernels. The central claim is that asset pricing routinely constructs, transforms, or selects probability measures (via discounting, numeraire normalization, marginal utility weighting, entropy penalization, calibration, or information conditioning) so that prices equal expectations under the adjusted measure, rather than estimating physical probabilities alone; the paper collects the associated formulas for state prices, Radon-Nikodym densities, Girsanov kernels, forward measures, coherent risk measures, benchmark pricing, and learned SDFs.","tokens_in":1933,"tokens_out":525,"duration_ms":32846,"significance":"If the synthesis and formula collection are accurate, the review offers a coherent organizing narrative around measure selection that is already implicit in standard references (Duffie, Cochrane) but here made explicit across the full historical arc to modern ML kernels. Its main value is pedagogical and referential: the explicit compilation of landmark formulas and the continuity argument from martingale methods to information-adjusted forecasting provide a compact reference that could aid teaching and literature navigation. No novel theorem or empirical result is claimed, so significance rests on the clarity and representativeness of the historical selection rather than on new technical content.","major_comments":[],"minor_comments":[{"comment":"The abstract and introduction list 'text-, attention-, and sentiment-based probability transformations' as recent extensions; a short dedicated paragraph or subsection clarifying their precise relation to the preceding martingale, numeraire, and SDF frameworks would improve readability for readers unfamiliar with the ML literature.","section":null},{"comment":"Notation for Radon-Nikodym derivatives and Girsanov kernels is introduced in multiple historical sections; a consolidated notation table or consistent symbol choice across the formula collection would reduce the risk of reader confusion.","section":null},{"comment":"The review cites canonical papers but does not include a brief discussion of scope limitations (e.g., omission of certain post-2010 continuous-time incomplete-market selection criteria); adding one sentence on selection criteria would make the narrative boundaries explicit without altering the central theme.","section":null}],"recommendation":"accept","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the positive assessment of the manuscript, the accurate summary of its scope, and the recommendation to accept. We appreciate the recognition of its potential pedagogical and referential value in tracing the role of probability measures across the historical development of asset pricing.","responses":[],"tokens_in":1429,"tokens_out":70,"duration_ms":17000,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper is a review that walks through the history of probability measures in asset pricing, starting with Bachelier and Arrow-Debreu state prices, moving through Black-Scholes risk-neutral valuation, Harrison-Kreps martingales, Delbaen-Schachermayer, SDFs from Hansen-Jagannathan and Cochrane, numeraire changes, and ending with recent ML-based kernels and information adjustments. It gathers formulas for state prices, Radon-Nikodym densities, Girsanov transforms, and related objects.\n\nIt does a reasonable job of organizing the material around the single theme that pricing routinely involves constructing or selecting measures rather than estimating physical probabilities. That framing is already standard, but pulling the formulas and milestones into one narrative can be convenient for someone learning the connections.\n\nThe soft spots are exactly what the abstract signals: no new results, no fresh derivations, and the historical selection is the usual canonical list with the usual risk of interpretive emphasis. The central claim does not rest on any fragile assumption because it is descriptive rather than predictive. If the full text accurately reproduces the cited formulas and does not introduce errors in the summaries, the technical content should hold up.\n\nThis is for readers who want a consolidated historical overview or a quick reference to the formulas. It is not aimed at specialists already fluent in the area. I would send it to peer review as a survey paper if the writing is clear and the citations are complete; it does not need to be rejected outright just because it is expository.","headline":"This is a straightforward historical review collecting standard results on probability measures in asset pricing with no new derivations or claims.","tokens_in":2386,"tokens_out":375,"would_cite":false,"duration_ms":27818,"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":"Asset pricing constructs transformed probability measures so prices equal expectations after discounting or utility weighting.","keywords":["asset pricing","probability measures","risk-neutral valuation","stochastic discount factor","pricing kernel","martingale measure","change of numeraire","incomplete markets"],"falsifier":"Identification of a major omitted foundational paper that alters the sequence from Arrow-Debreu state prices through martingale measures to modern learned kernels, or market data in which untransformed physical probabilities price assets as accurately as the reviewed transformations.","tokens_in":2738,"feed_emoji":"","tokens_out":612,"duration_ms":24622,"temperature":0.7,"pith_summary":"The review traces how asset pricing moved from physical probabilities to specially constructed or transformed measures that let prices appear as discounted expectations. It shows successive refinements through state prices, risk-neutral measures, numeraire changes, stochastic discount factors, and data-driven kernels. A sympathetic reader sees this as the reason observed prices deviate from real-world odds in systematic ways. The paper unifies these steps by collecting the formulas that implement each transformation. Readers care because the approach explains why direct statistical estimation of physical probabilities rarely recovers market prices.","feed_headline":"Pricing uses adjusted probabilities to match market values","feed_subtitle":"Review shows asset pricing selects measures via discounting, numeraire changes, and utility weights rather than real-world odds.","key_machinery":"Change of probability measure, implemented through Radon-Nikodym densities, Girsanov transformations, or pricing kernels, that converts physical probabilities into a measure under which prices equal conditional expectations.","core_discovery":"Asset pricing is not merely an exercise in estimating physical probabilities. Instead, pricing theory constructs, transforms, or selects probability measures so that market prices can be represented as expectations after discounting, numeraire normalization, marginal utility weighting, entropy penalization, calibration, or information conditioning.","pith_inferences":["New pricing models should state their measure transformation explicitly instead of defaulting to physical probabilities.","The framework links asset pricing to decision theory by treating the choice of measure as an integral modeling step.","Empirical work could test whether entropy-penalized kernels outperform purely data-driven ones during high-volatility periods."],"forward_implications":["Risk-neutral valuation prices derivatives as discounted expectations under the equivalent martingale measure.","Stochastic discount factors impose volatility bounds on admissible pricing measures via the Hansen-Jagannathan relation.","Incomplete markets require explicit selection criteria such as entropy minimization among equivalent measures.","Machine learning methods can learn pricing kernels directly from text or sentiment data while preserving earlier frameworks.","Information-adjusted forecasts complement rather than replace numeraire and SDF approaches."],"fun_headline_variants":["Probability measures constructed for asset pricing not physical estimates","Asset pricing selects transformed measures for market representation","From Arrow-Debreu to SDFs: evolving probability measures in finance","Market values expressed as expectations under selected probability measures"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The selected landmark contributions form a representative sequence of key developments without major omissions or interpretive bias.","fun_headline_variants_meta":{"raw":{"variants":["Probability measures constructed for asset pricing not physical estimates","Asset pricing selects transformed measures for market representation","From Arrow-Debreu to SDFs: evolving probability measures in finance","Market values expressed as expectations under selected probability measures"]},"model":"grok-4.3","cost_usd":0.00851,"raw_usage":{"total_tokens":3870,"prompt_tokens":717,"num_sources_used":0,"completion_tokens":61,"cost_in_usd_ticks":85099500,"prompt_tokens_details":{"text_tokens":717,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":3092,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":717,"tokens_out":61,"duration_ms":35465,"temperature":1.0,"reasoning_tokens":3092,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-29T13:55:47.120186+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Identification of a major omitted foundational paper that alters the sequence from Arrow-Debreu state prices through martingale measures to modern learned kernels, or market data in which untransformed physical probabilities price assets as accurately as the reviewed transformations.","supporting_citations":[],"review_version":1}