{"id":"42dc8287-3fe9-4197-95ac-37cd6a04e01f","arxiv_id":"2607.24680","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":4,"one_line_summary":"A parsimonious stretched-exponential quantile splice parameterizes risk-neutral distributions and calibrates accurately to hundreds of thousands of SPX implied-volatility curves, including concave shapes.","lead":"The paper introduces a stretched-exponential quantile-splice model that parameterizes the risk-neutral distribution with a handful of interpretable knobs and fits a wide range of S&P 500 implied-volatility shapes, including local concavity. Stable tenor patterns in the fitted knobs are meant to support arbitrage-free surface interpolation and local-volatility dynamics.","discovery_kind":"new_method","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"The \"no static arbitrage\" claim rests on Prop. 6.1's integral condition, but no interpolant θ(τ) is ever constructed or checked against it — and the condition is a continuum of inequalities that generic smooth interpolation will not automatically satisfy.","rationale":"The reader's weakest_assumption is, in my independent read, the correct load-bearing point, and my critique sharpens rather than replaces it: Prop. 6.1 imposes infinitely many simultaneous inequalities on θ′(τ), existence of a feasible direction is never shown, and the short-tenor regime (where the initial condition bites and the fitted parameters move fastest) is the likeliest violation site. I checked whether a different concern might supersede this one: the calibration evidence (residuals vs. spreads across curve/quotation/tenor/moneyness views, §4.3) is internally consistent and the outlier analysis (§4.4) is candid about model limits; the propositions I traced (2.1, 3.4, 3.5, 5.1, 5.4, 5.5, 6.1, 6.2) follow from standard arguments (Weibull moments, change of variables, Rolle's theorem, Fokker–Planck integration) with no visible gaps; the hybrid model's λ̂ > 1 fits would need condition (5.1) verified numerically, but that is a secondary, checkable detail rather than a load-bearing gap. The reader already conditioned acceptance on exactly this missing artifact (explicit feasible θ(τ) example), so no verdict change is warranted: UNCHANGED, i.e., keep CONDITIONAL. The proposed test is decisive in both directions and is computable with the paper's own machinery.","tokens_in":35011,"tokens_out":2139,"duration_ms":85667,"concrete_test":"Take one calibration day (e.g., 2024-01-02, the 54 tenors of Fig. 5). Build an explicit θ(τ): shape-preserving cubic (PCHIP) per parameter through the fitted points, with λ(τ) anchored to λ(0)=0. On a grid of (τ, u) (e.g., 2000 × 512), numerically evaluate J(τ,u) = ∫₀ᵘ e^{Q(v;θ(τ))} ∇_θ Q(v;θ(τ))·θ′(τ) dv via the same quadrature used for pricing. If max J > 0 anywhere, the naive interpolant admits calendar arbitrage; then re-solve the interpolation with J(τ,u) ≤ 0 as a constraint (or penalty) and re-measure curve-wise RMS residuals (eq. 4.5) at listed tenors. If constrained interpolation keeps residuals inside the spread bands of Fig. 6, the \"without static arbitrage\" claim is substantiated; if residuals materially exceed spreads or no feasible θ′ exists at some τ, the claim must be weakened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest claim has two parts: (i) accurate curve-wise calibration (§4, Figs. 4–6) and (ii) stable tenor patterns \"enabling term structure interpolation and dynamic process construction without static arbitrage\" (Abstract, §6). Part (i) is well evidenced. Part (ii) is where the load is least supported. Prop. 6.1 requires, for all τ ∈ 𝒯 and all u ∈ (0,1) simultaneously, ∫₀ᵘ e^{Q(v;θ(τ))} ∇_θ Q(v;θ(τ))·θ′(τ) dv ≤ 0. This is not a smoothness condition on θ(τ): it is an uncountable family of half-space constraints on the direction θ′(τ) at every tenor. Nothing in §6 shows a feasible direction exists at any τ, let alone that the visually smooth paths in Figs. 5, 12–14 admit one. The inference \"clear pattern ⇒ existence of a feasible term structure\" (§6, first paragraph) is precisely the unproven step. Two specific stress points: (a) near τ→0 the initial condition forces λ(τ)→0 with Q(u;θ(0))=0, while Fig. 5 shows γ±₁, γ±₂ varying most rapidly exactly at short tenors — the regime where the integrand's sensitivity to θ′ is largest; (b) moneyness-dimension arbitrage is ruled out only because each fixed-τ splice is a valid RND, but the calendar condition couples all seven components of θ′, so independent per-parameter spline interpolation (the natural reading of \"interpolation\") has no reason to respect it. Since this claim is one of the paper's two headline deliverables and is currently supported only by visual stability, the reader's CONDITIONAL verdict and its named condition are well placed.","agreement_with_reader":"agree"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The paper introduces a family of risk-neutral distributions built by splicing stretched-exponential (Weibull) quantile segments with C¹ matching at knots, closed by an additive risk-neutralizing shift. The three-segment instance has seven free parameters (scale ς, two knots u±₂, four shapes γ±₁, γ±₂) with tail segments fixed. The author develops the construction pedagogically (logistic → beta-logistic → stretched logistic → splice), proves risk-neutralization (Prop. 3.4) and wing asymptotics (Prop. 3.5) from Lee's moment formula and Benaim–Friz, and calibrates independently each of 25,770 SPX/SPXW implied-volatility curves over 2024–2025 (~5M option pairs). Curve-wise RMS implied-vol residuals are reported to lie mostly within RMS half bid–ask spreads; outlier curves with strong concavity are handled by a four-segment model or a hybrid affine combination of two three-segment RNDs, whose feasibility range is characterized (Props. 5.1, 5.4). Section 6 states the calendar no-arbitrage condition (Prop. 6.1) for a parameter term structure θ(τ) and gives Dupire local-volatility formulas (Prop. 6.2, Cor. 6.3).","tokens_in":35508,"tokens_out":3290,"duration_ms":99412,"significance":"The quantile-splice construction is genuinely neat: it exploits the convex-cone structure of quantile functions to obtain a parsimonious, interpretable parameterization with direct, localized control over implied-vol shape, explicit integrability conditions, and closed-form wing asymptotics via Propositions 3.4–3.5 — properties most flexible smile parameterizations lack. The empirical study is large and carefully sliced (curve-, quotation-, tenor-, and moneyness-wise residuals against spreads), and the handcrafted calibration walkthrough (§4.1) substantiates the interpretability claim. The affine-combination feasibility result (Prop. 5.1) and the zero-crossing counting argument (Prop. 5.5, generalizing Glasserman–Pirjol) are clean, correct-looking contributions. If the calibration results hold, this is a practically useful model with an unusually transparent parameter-to-shape mapping, and it would be a solid reference for event-driven concave smiles. The weakest link is the advertised bridge to arbitrage-free surfaces and local-vol dynamics, which is asserted rather than demonstrated.","major_comments":[{"comment":"The abstract and §6 claim that the stable tenor patterns in the fitted parameters \"enable term structure interpolation and dynamic process construction without static arbitrage.\" This is one of the paper's two headline deliverables, and it is not supported. Proposition 6.1's no-arbitrage condition — for all τ ∈ 𝒯 and all u ∈ (0,1), ∫₀ᵘ e^{Q(v;θ(τ))} ∇_θ Q(v;θ(τ))·θ′(τ) dv ≤ 0 — is a continuum of half-space constraints on the direction θ′(τ) at every tenor, coupling all seven parameter components. Smoothness of the discrete paths {θ̂_t(τ)} in Figs. 5 and 12–14 is neither necessary nor sufficient for feasibility, and no interpolant θ(τ) is ever constructed or checked against the condition. Independent per-component spline interpolation (the natural reading of \"interpolation\") has no reason to respect it. The short-tenor regime is the sharpest stress point: the initial condition Q(u;θ(0)) =","section":null}],"minor_comments":[],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The useful core here is a three-segment stretched-exponential quantile splice (seven free parameters after fixing tails) that gives localized, interpretable control over IV shape, including mild local concavity, and actually fits almost all of 25k SPX/SPXW curves from 2024–25 with residuals inside spreads.\n\nWhat is new is the specific construction: piecewise Weibull quantiles glued C1, risk-neutralized by the usual additive shift, with explicit wing formulas and a hybrid affine extrapolation for the few extreme W/S shapes. The supporting propositions (risk neutralization, wings, feasible affine combos, local-vol expressions) are standard and carefully written. The empirics are the real strength—hand-crafted intuition, full-day surfaces, residual histograms vs bid–ask, tenor- and moneyness-wise bands, and outlier analysis with four-segment and hybrid fixes. Parameters do show stable tenor patterns (scale grows, knots sit near 0.03/0.98, short-tenor γ1 small). That is genuine engineering value for vol-surface work.\n\nSoft spot, in proportion: the abstract and §6 claim that those patterns “enable term-structure interpolation and dynamic process construction without static arbitrage.” Prop. 6.1 correctly states the integral calendar condition on θ′(τ), but no interpolant is ever built or checked against it. Visual smoothness of the discrete paths is not the same as existence of a feasible direction at every τ (especially short tenors where the γ’s move fastest). Moneyness arbitrage is automatic per slice; calendar is not. So the second headline deliverable is only half-supported. Minor additional limits: closed Bloomberg data, no code, no quantitative baselines against mixtures or SSVI-style params.\n\nThis is for people who build or recalibrate equity IV surfaces and want an interpretable RND that handles event-day concavity without jumping to high-component mixtures or nets. Math and residual evidence are solid enough that a serious editor should send it to referees; the dynamic claim just needs an explicit feasible θ(τ) example or a quieter rewrite. I would read the revision and probably cite the splice construction.","headline":"Solid quantile-splice RND with real SPX calibration power and clean math; the arbitrage-free dynamic claim is asserted from visual parameter stability, not shown.","tokens_in":31923,"tokens_out":529,"would_cite":true,"duration_ms":11232,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91G20","60E05","91G60"],"pacs":[],"model":"grok-4.5","headline":"A seven-parameter quantile splice fits a quarter-million SPX volatility curves and yields stable tenor paths.","keywords":["option pricing","implied volatility","risk-neutral quantile","quantile splice","stretched exponential","local volatility","static arbitrage","SPX options"],"falsifier":"Construct an explicit continuous interpolant of the calibrated seven-parameter paths on a multi-tenor SPX surface, compute the associated Dupire local volatility, and check whether the resulting put prices remain free of calendar-spread arbitrage (partial_tau p >= 0) across a dense grid of tenors and strikes.","tokens_in":31538,"feed_emoji":"📈","tokens_out":815,"duration_ms":15935,"temperature":0.7,"pith_summary":"Modern short-dated equity options often show implied-volatility curves that are not simply U-shaped: they can be W-shaped, inverse-U, or S-shaped, with local concavity near events. Classical stochastic-volatility or mixture models either need many components or lose interpretability to capture those shapes. This paper parameterizes the risk-neutral quantile function itself by splicing stretched-exponential pieces. A three-segment version with seven free parameters gives direct, localized control over level, wing slope, and central convexity. On two years of SPX and SPXW data the model calibrates essentially every curve to within the bid–ask spread; the fitted parameters form smooth patterns across tenor. Those patterns let one interpolate a whole surface and build a local-volatility dynamics free of static arbitrage. The practical payoff is a single, parsimonious object that both prices today’s book and seeds tomorrow’s process.","feed_headline":"Seven parameters fit a quarter-million SPX volatility curves","feed_subtitle":"A quantile splice captures U-, W- and S-shapes and yields stable tenor paths for arbitrage-free dynamics","key_machinery":"The stretched-exponential quantile splice: the risk-neutral quantile is built piecewise from shifted Weibull quantiles joined C1 at free knots; scale, knot locations and shape exponents give direct, localized control over level, wing slope and central curvature of the implied-volatility curve.","core_discovery":"A risk-neutral stretched-exponential quantile splice with three segments and seven free parameters (global scale, two interior knots, four shape exponents; tails fixed) reproduces the full range of observed SPX implied-volatility shapes—including local concavity—and calibrates accurately to roughly 25 770 curves spanning 2024–2025, with residuals typically inside market spreads and with tenor-stable parameters that support arbitrage-free term-structure interpolation.","pith_inferences":["The same splice construction could be applied day-by-day to single-name or crypto option books where event-driven concavity is even more pronounced.","Because the parameters are directly interpretable, they may serve as low-dimensional state variables for a reduced-form stochastic model of the surface itself.","Hand-tuning the seven parameters without an optimizer already produces usable fits, suggesting the model could be used for rapid scenario generation on a trading desk."],"forward_implications":["A single seven-parameter object prices the bulk of the SPX book to market-spread accuracy.","Stable tenor patterns of the seven parameters allow direct interpolation of entire implied-volatility surfaces.","The same parameter paths seed a Dupire local-volatility diffusion that matches the surface by construction and rules out static arbitrage.","Four-segment or hybrid (affine combination) extensions cover the rare extreme-concavity outliers without leaving the same family.","Wing asymptotics are set a priori by the fixed tail exponents, removing an extra free parameter while still matching observed short-dated wings."],"fun_headline_variants":["Seven parameters fit 250k SPX vol curves with local concavity","Quantile splice captures U-W-S vol shapes across quarter-million curves","Seven-param risk-neutral model calibrates full SPX implied-vol range","Stable seven parameters enable arbitrage-free SPX tenor interpolation","Three-segment splice fits 25770 SPX curves inside market spreads"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"That the visually smooth discrete parameter paths across listed tenors can be turned into a continuous term structure that satisfies the integral no-arbitrage condition for every tenor and moneyness.","fun_headline_variants_meta":{"raw":{"variants":["Seven parameters fit 250k SPX vol curves with local concavity","Quantile splice captures U-W-S vol shapes across quarter-million curves","Seven-param risk-neutral model calibrates full SPX implied-vol range","Stable seven parameters enable arbitrage-free SPX tenor interpolation","Three-segment splice fits 25770 SPX curves inside market spreads"]},"model":"grok-4.5","effort":"low","cost_usd":0.004296,"raw_usage":{"total_tokens":1175,"prompt_tokens":634,"num_sources_used":0,"completion_tokens":81,"cost_in_usd_ticks":42964000,"prompt_tokens_details":{"text_tokens":634,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":460,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":634,"tokens_out":81,"duration_ms":8808,"temperature":1.0,"reasoning_tokens":460,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T08:03:18.512407+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Construct an explicit continuous interpolant of the calibrated seven-parameter paths on a multi-tenor SPX surface, compute the associated Dupire local volatility, and check whether the resulting put prices remain free of calendar-spread arbitrage (partial_tau p >= 0) across a dense grid of tenors and strikes.","supporting_citations":[],"review_version":1}