Papers

Ethereum sells data-availability capacity to rollups through "blobs" priced by a controller: the blob fee rises and falls exponentially with cumulative usage above or below a target, and no auction sets it. We measure what this mechanical price costs. The Pectra upgrade of 7 May 2025 doubled blob capacity on a date fixed weeks in advance, while the market was congested, and the blob fee fell from 0.82 mETH per blob (about USD 1.40) to its 1-wei floor. Rollups' blob purchases rose 24%. The 95% interval, built from the same estimator at 516 placebo dates, runs from 8% to 40%, so removing the entire price raised purchases by at most 40%. The implied elasticity at the pre-shock price is -0.19, and no value below -0.31 is consistent with the interval. The estimate includes any rise in L2 activity that coincided with the fork, so it bounds the price response from above. With demand this inelastic, the controller works mainly as a transfer. Rollups paid USD 14.8 million in blob fees under the first blob regime, 67% of it in two quarters when capacity was full and any market-clearing mechanism would have collected comparable scarcity rents. Where capacity was slack, the reserve price introduced in December 2025 cost at most 2.6 ETH (USD 8,039) in deadweight loss under the most elastic demand the data allow. Mempool records for 5.4 million blob transactions show that when the fee sits at its floor, capacity is slack: transactions miss about one block and are almost never priced out.

A path indistinguishable from its own reverse carries no arrow of time, so time irreversibility is a necessary condition for a directed mechanism to be readable from trajectories alone. In finance it has been measured only on scalar series such as returns and volatility. We define the irreversibility rate of a coarse-grained limit order book path as the forward-to-reverse log-density ratio per event, and estimate it from three weeks of Nasdaq market-by-order data with an order-invariant MLP and a temporal convolutional network that receive identical blocks. For NVDA, the order-invariant and order-sensitive estimates are 0.036 and 0.054 nats per event, respectively. Their difference, the ordering increment, is 0.018. A surrogate preserving the coarse one-step transition table recovers 30 percent less irreversibility than the observed paths. The increment is positive for all twelve symbols on every held-out daytime date, and for NVDA in each session window. It changes more than tenfold with the coarse-graining scale. Neither a linear score nor any single feature group reproduces the full fitted rate. A queue-reactive simulator calibrated on the same days matches the market's ordered rate, 0.052 against 0.054, yet its ordering increment is 0.0006 against 0.018, building temporal asymmetry out of block composition instead of block order. We propose the ordering and scale profile of path irreversibility as a benchmark for order book simulators. Code and audit materials are available at https://github.com/hjyeo30/lob-irreversibility.

EIP-4844 gave Ethereum rollups a separate fee dimension for data, governed by a multiplicative update rule that targets three “blobs” per block. Whether this target is reached depends on how many independent, self-interested rollups choose to post data, a question that closed-form models address only for a single agent under exogenous prices or in a static equilibrium. We propose a calibrated multi-agent reinforcement learning environment for the blob fee market in which the blob base fee is endogenous, set by the EIP-4844 rule on the aggregated decisions of heterogeneous rollups fit to post-Dencun Ethereum data. After confirming that the environment reproduces the known single-agent analytical optimum, we show that independent learners, with no convergence guarantee, still settle near the target without coordination ($3.32\pm0.01$ blobs per block over five seeds). Over a fiftyfold demand range ($10\times$ to $500\times$), supply stays at the target from $10\times$ to $200\times$ and rises above it at higher demand, where the fee can no longer hold it. Posting frequency increases with rollup size, an emergent analog of the size threshold predicted by prior theory. A costless layer-1 calldata outside option pushes the fee down by roughly an order of magnitude while aggregate supply stays at the target, so the fee rule acts as a negative-feedback controller on decentralized supply.

Temporal point processes are used both to measure market mechanisms, for example how fast prices absorb news and how much of the reaction is the market responding to itself, and to predict the next event. We show that one algebraic object, the phase-type (Erlang-chain) state space, serves both purposes with exact likelihoods. For measurement, we build a multiscale phase-type Hawkes process whose parameters all enter linearly. Its log-likelihood is concave, and we compute the global maximum together with a Fenchel duality-gap certificate. A Markov embedding gives the response to an exogenous shock, echoes included, in closed form, and this response is stable if and only if the spectral radius of the branching matrix is below one. Applied to six markets around 670 scheduled macroeconomic releases (2022-2026) with a built-in placebo test, the model shows that the direct reaction to CPI, payrolls, PPI and the FOMC statement is absorbed in about a second, while the market's own echoes stretch half of the response to 16-137 s. For prediction, we introduce EPT-X, a neural point process that adds a history-dependent mixture-of-Erlang renewal channel to a linear Hawkes backbone. Its intensity and compensator remain in closed form, its states follow an affine recursion that trains with a parallel scan, and it can start from the certified fit of the measurement model. EPT-X has the best test likelihood on all five LOBSTER order-book streams and on Retweet and Taobao, and is level with the best published model on StackOverflow. We also find that the EasyTPP implementation of IntensityFree clamps short inter-event times before scoring, and that the gaps of the EasyTPP Amazon data are uniform on two fixed intervals.

This paper studies exit timing for SPX short-straddle portfolios using a two-stage supervised and reinforcement-learning framework. Historical episodes first generate ex-post optimal hold/exit labels for training a multilayer perceptron (MLP) using only contemporaneously available state variables at deployment. The trained MLP then initializes reinforcement-learning actors that are fine-tuned with a risk-aware trading reward. Out-of-sample evaluation over 2024-2026 shows that the supervised MLP substantially outperforms passive holding and representative fixed-time and profit-and-loss-based exit rules. Across seven independent seeds, MLP-initialized Actor-Critic averages 37.54% cumulative return and a Sharpe ratio of 1.705, compared with 35.01% and 1.651 for the supervised MLP. Actor-Critic trained from scratch averages only 0.18% return. These findings show that oracle-guided supervised learning provides a stable foundation for reward-based policy fine-tuning in a sparse and path-dependent option-exit problem.

FLEX options are customized exchange-traded contracts allowing participants to choose nonstandard strikes, expirations, exercise styles, and settlement terms while retaining central clearing. We investigate whether single-name FLEX activity leaves measurable footprints in regular listed-option markets and underlying equities. Using a linked dataset of FLEX consolidations, daily open-interest files, listed-option chains, equity returns, and corporate actions over 250 trading days (March 2025-March 2026) across 41 underlyings and 53 event dates, we document three primary empirical findings: (1) Matched-control difference-in-differences estimates reveal a robust post-event increase in listed-option open interest (beta = 0.3061, p = 0.0042, Benjamini-Hochberg p = 0.0168). (2) FLEX activity is strongly selected into high implied-volatility (IV) environments, concentrated in mid-price underlyings (date-fixed-effect ATM IV premium beta = 0.1235, p = 0.0030). (3) Event-day IV-minus-realized-volatility spreads predict execution-stressed short-straddle returns (beta = 0.0366, p = 0.0100). Short-lived equity return effects (CAR_0 difference of -0.85%, p = 0.0442) do not persist over 1-to-5-day horizons. Overall, FLEX activity serves primarily as an observable marker of institutional volatility demand and open-interest positioning rather than a directional stock-return predictor.

When a trader executes a large order, the relevant question is not what will the price be? but what would the price have been had I not traded? Market impact is thus a causal quantity, yet classical models (Almgren–Chriss, propagators) regress price on flow, which is biased whenever informed flow moves prices through information as well as trading. We present a proof-of-concept Causal Market Simulator: a continuous-time causal prior-data fitted network (PFN) trained on a synthetic market prior in which a hidden informational flow confounds the flow–price relation by construction, with confounding strength an exact per-episode knob. On this prior, the naive estimator's bias follows a closed form we confirm empirically (up to 68% overestimation), while the PFN tracks the unbiased true-parameter oracle at high confounding. Its do-ablated twin, the same model with the intervention inputs removed, collapses onto the pre-window mean. Zero-shot, the model responds at the ground-truth scale to metaorders in an ABIDES limit-order book ($+0.38$ pre-onset standard deviations against seed-replay counterfactuals), and its intervention response replicates on about 2,300 detected flow bursts in BTC/ETH/SOL perpetuals and 3,800 on Nasdaq equities, evidence consistent with, though not proof of, causal transfer. Off-prior the response degrades gracefully, and split-conformal recalibration repairs interval calibration. Transfer hinges on encoding the metaorder as added flow rather than a clamped price level, not on more training. A recurrent state-space variant closes the remaining shape gap: its impact accrues over execution duration where the transformer's stays flat, raising the correlation with measured metaorder impact from $+0.68$ to between $+0.92$ and $+0.96$, at the cost of weaker on-prior deconfounding that traces to the training recipe rather than to recurrence.

Structural price diffusivity explains many empirical regularities of market impact including the “square-root law” and its crossover to a linear regime for low trading rates \citep{bonart2026diffusive}. One of its central predictions is that an information-neutral trading strategy generates a diffusive impact state. We microfound this result in an economy with a trader and many arbitrageurs who progressively eliminate predictable returns. Each arbitrageur observes realized returns and a private signal of one component of the fundamental return. As aggregate private information becomes complete and under suitable convergence to a stationary limit, the trader's impact law is of the form $j=U\cdot w$, where $w$ is the innovation in the trading schedule and $U$ is causal all-pass. Impact returns are therefore white, even though no individual arbitrageur can reconstruct the underlying trading strategy. We then investigate the economic meaning of the impact phase. We argue that the $U$ which has minimum distance from the filter $L$ generating the trade flow is an especially interesting candidate: Trade flow is then minimally distorted under impact and arbitrage, and it always guarantees positive impact costs. Under a trader flow generated by a biexponential $L$ (the simplest form allowed in our model) an interesting result emerges: Impact decline is confined to a narrow band of $50\%$–$60\%$, lower than some empirical estimates and quite consistent with others. More complicated flow models can lead to different decays. Finally, we argue that in real markets, the impact phase is probably somewhat “distorted” in its long-term tails which allows for a full relaxation of the impact propagator. The effect is weak, long-term mean-reversion of the impact state.

We provide an event-level Hawkes microfoundation for a multitype inverse-Gaussian stochastic clock. We show that the event counts and integrated intensities of nearly critical multivariate linear Hawkes processes converge jointly to a multivariate pure-jump subordinator when reproduction delays have finite mean and immigration is balanced so that rare large families remain visible. The dependence among the limiting coordinates is inherited from microscopic cross-excitation, and the limit admits a Brownian additive-field first-passage representation. We call this process the multitype inverse-Gaussian subordinator. Its diagonal case consists of independent classical inverse-Gaussian subordinators, with the univariate model as a further special case. The square-root specialization recovers the inverse-Gaussian clock obtained at the boundary of hyper-rough square-root models by Abi Jaber–Attal–Rosenbaum (Ann. Appl. Probab. 36(4): 3635–3660, 2026). Our cluster proof exposes the underlying mechanism: finite-variance near-critical branching creates rare but macroscopic families, while their internal timing disappears on the observation scale, so each family becomes a jump. We also clarify the relation with the multivariate Hawkes scaling theory of Xu (arXiv:2412.14459): its diagonal-atom condition yields the diagonal specialization, whereas the genuinely coupled limit requires a separate uniqueness argument. In addition, under a common tilted-stability condition, we replace the high-intensity assumption in that theory by a weaker accumulated-activity condition and provide the Riccati identification needed when the potential has an atom. We strengthen convergence of the count and compensator and extend the scalar conclusion to finite-variance age-dependent branching clusters.

As option markets grow and AI advances, agentic systems for option trading are gaining increasing attention. Language-model-based agents can reason over contextual information such as news, but option trading presents a particularly challenging decision problem: a single stock can have thousands of contracts, and the agent must decide both which contracts to trade and how to combine them. Existing approaches often sidestep this complexity by restricting the policy to a fixed strategy structure, such as a straddle, limiting their ability to switch strategies as market conditions change. We present SOTA (Stock Options Trading Agents), an agentic trading framework for structured option-strategy selection. SOTA abstracts the large option universe into strategy-level decisions while deterministic resolvers handle portfolio implementation. We develop SOTA by post-training Qwen3.8-27B with supervised fine-tuning followed by reinforcement learning. SOTA is evaluated on options on nine large-cap U.S. equities and SPY against rule-based and machine-learning strategy selectors in the same trading environment. Over a six-month out-of-sample period, SOTA earns an 18.3% total return with a Sharpe ratio of 1.60 and a maximum drawdown of 8.96%. We also document an asymmetric role of news: news improves frontier-teacher trajectories, but retaining news during reinforcement learning reduces out-of-sample return from 18.3% to -2.7%.

The square-root impact law (SRIL), $I = Y\sigma\sqrt{Q/V}$, bundles two facts that a single mechanism must explain at once: a shape (impact proportional to square-root of traded volume $Q$) and an amplitude ($Y=O(1)$, independent of the participation rate $\varphi$). Bonart has recently proposed an elegant solution: if realized and counterfactual prices are both diffusive, information-neutral impact must have white increments and the SRIL follows without the wart. We reformulate and simplify his argument, and foreground the ingredient he himself regards as essential — that the market whitens the stationary stream of each participant's metaorders, not any isolated one. The Lillo–Mike–Farmer model satisfies his premises, yet gives an isolated metaorder the super-square-root impact $\varphi^{(1-\gamma)/2}Q^{(1+\gamma)/2}$, $\gamma\in(0,1)$, because single metaorders (i.e. not part of a stream) are statistically invisible and thus priced mechanically. We then construct an explicit multi-agent propagator that realizes Bonart's whitening, and show it reproduces the clean SRIL precisely when the market can attribute trades to their issuer. But correct re-attribution of anonymous trades is a formidable task, which requires markets to behave as implausibly efficient signal processing machines. The stream hypothesis makes a sharp, falsifiable prediction — isolated metaorders, well separated in time, should not obey the SRIL — that does not seem agree with empirical data known to us. Independently, a volume-conservation argument singles out the square-root law and lays bare the step the diffusivity route must assume: diffusivity fixes the variance of impact — an additive, stationary, age-blind quantity — whereas the SRIL is a law for the mean impact, which marginally decays as $1/\sqrt{\text{age}}$ and therefore demands an origin of time.

The growing importance of intraday trading in Europe, driven by the increasing penetration of renewable energy sources, has led to higher volatility and periods of market stress. Understanding how order flow behaves under varying liquidity conditions requires models that adapt to the state of the market. This paper proposes a multivariate state-dependent Hawkes process to model the flow of orders across different liquidity regimes in the intraday continuous electricity market. The observed market states are identified using a proposed Liquidity Stress Index (LSI), which quantifies stress using the bid-ask spread in the limit order book, order book volume, and mid-price volatility. Transitions between states induce changes in the intensity parameters, resulting in piecewise dynamics that reflect regime-specific order-flow behavior. The model is calibrated on January-March 2024 Dutch intraday continuous cross-border intraday (XBID) hourly order data from EPEX SPOT. The estimated dynamics show that order flow is strongly self-exciting and near-critical, that executed trades trigger new same-side orders while the reverse effect is negligible, and that liquidity stress mainly strengthens transaction self-excitation rather than reshaping the cross-side structure. Robustness checks over the LSI composition and the state partition show that the results are driven by the regime construction itself: alternative index compositions fit within a narrow band, three states fit better than two, and removing the states clearly degrades the fit relative to a single-regime Hawkes benchmark. The results highlight both the value of conditioning order-flow models on liquidity stress and the limits of a single-exponential specification when event frequencies are highly imbalanced.

We give a unified analytic account of a finite bid–ask spread in a two-field reaction–diffusion order book. The model retains separate bid and ask densities in operational time, with diffusion, cancellation, reaction and external order creation. On the symmetric equal-coefficient branch, the imbalance field determines the reaction price, while the total standing density carries the spread geometry. A market-making source is introduced by withholding order placement over a finite interval around the reaction boundary, representing round-trip replenishment displaced from the quote. Under a separated, quasistationary and weak-overlap approximation, transport penetrates inward from the placement edges over the cancellation length, giving the quoted spread as the sub-threshold core of the placement-source-free interval. Market orders remove standing density at the observed quote, whereas replenishment is delivered outside it and must diffuse back across the penetration depth. We derive the resulting delayed and attenuated quote-level response in operational time. On that branch, finite spread and price impact arise from distinct response sectors, while their calendar-time appearance requires a separate observation clock.

Trade-sign errors can change measured trading costs even when classification accuracy is high. We validate the side of Polymarket's public trade prints against the taker leg of each print's on-chain settlement. On twelve selected days between April and August 2026, spanning both exchange generations, 92.1% to 100.0% of prints match a settled taker leg, with exact side and token agreement on all 24.5 million matched pairs. Mint-and-merge settlement makes that taker leg essential: pooling maker and taker legs changes the measured buy share. Signing the full cached tape before settlement selection yields 16.6 million prints signed by every rule; equal-day balanced accuracy is 0.942 for Lee-Ready, 0.768 for the tick test and 0.670 for retrospective bulk volume classification. On 16.4 million identical eligible fills, Lee-Ready raises effective spread by 0.429 cents per share under equal-fill weights; its realised-spread difference is -0.209 cents. Under share-volume weights, taker five-minute midpoint impact is 0.689 cents, while tick and bulk classifications give -0.374 and -0.119 cents. That aggregate sign reversal disappears when tied print rows are excluded. Distortion depends on error-weighted signed outcomes, sample selection and weighting. Receipt-time ordering and unobserved future-quote age limit these delivered-quote accounting quantities; they do not identify causal impact or private information. Supporting venue and collector analyses provide descriptive diagnostics.

Market-regime models typically assume a finite set of discrete latent states. We examine whether high frequency limit-order-book dynamics exhibit distinct regime separation or apparent regimes result from discretising an underlying continuum, analysing deep limit-order book data for EURO STOXX 50 index futures across 987 clean trading days from 2022 to 2025 using 30, 45 and 60-second aggregation windows. Market states are represented by symmetric positive definite covariance matrices and analysed under Log-Euclidean and affine-invariant geometries; the resulting state cloud has an effective dimension of approximately 1.4. A causal anomaly layer filters statistical, calendar and rollover contamination, while label-free economic validation evaluates market-state separation independently of volatility-based proxy labels. A 17-method zoo spanning conventional and Riemannian representations, reinforced by full-cloud geometric certificates, consistently favours a continuum over a discrete regime structure. A dominant latent coordinate captures between 83.66% and 85.27% of variation in the covariance-state geometry and tracks VSTOXX without regime labels, remaining stable across years although its average level shifts with market conditions, indicating that conventional regimes are coarse quantisations. A divisive hierarchy spanning macro, macro-subregime, micro and micro-subregime tiers is evaluated using expanding-year walk-forward tests, with out-of-sample performance strongest at the macro tier while finer tiers fail to generalise; continuous fine-scale information nonetheless remains predictive, and its discretisation causes the performance loss. Forecasting performance peaks near ten minutes, reaching a maximum pooled R2_OOS of 0.7483. Overall, the latent stress coordinate defines a stable, low-dimensional geometric continuum that captures market stress and its predictive dynamics.

We study when to buy a share that will later be sold optimally, when the price follows a geometric multi-skew Brownian motion whose skew levels model support and resistance. The reward for buying is the exit premium of the liquidation problem solved in a companion paper. This premium is strictly $r$-subharmonic inside the exit continuation region, so buying is optimal only at levels where the price is pushed upward, and the entry problem reduces to a finite one whose value is an upper concave hull of finitely many points. With one support and one resistance, the optimal rule is to buy exactly at the support, in every exit regime, with an explicit value.

Financial LLM agents are often evaluated by comparing their end-to-end returns with those of a baseline and testing the paired difference against zero. This measures whether deploying the agent changes realized performance, but it does not isolate event-selection skill. An agent that frequently changes positions from flat to long can earn a positive paired return from an upward-drifting event pool even when it selects events at random. We propose the Agent Policy-Value Audit, which holds fixed the observed count of each ordered action-change type and randomly reassigns them across eligible events. The average payoff from these reassignments is the composition benchmark; the difference between observed deployment value and this benchmark is selection value. In semi-synthetic benchmarks based on real earnings-event returns, a zero-centered paired test falsely attributes passive exposure to selection skill in $11.6\%$ of no-skill replications, while the transition-matched audit reduces this rate to $5.3\%$. Applied retrospectively to 723 earnings events at 44 U.S. consumer-facing firms, the audit decomposes the agent's gross deployment value of $+15.2$ bps/event into a $+25.8$ composition benchmark and a $-10.6$ selection value. The agent does not detectably outperform matched random assignments. Financial-agent evaluations should report deployment value separately from event-selection value.

Flashblocks divide a block's priority gas auction into shorter sequential auctions that commit transaction order before the block is complete. We ask how this auction cadence affects bidding and competition among automated arbitrageurs, or searchers. In July 2025 Base replaced a single 2 s auction with ten 200 ms auctions. We use this change to estimate the searcher response from on-chain data alone. On an address-day panel of 3,032 searchers and 8,053 activity-matched controls, a difference-in-differences design estimates a 0.187 gwei fall in the effective priority fee (59% of the searcher pre-period mean). The share of searcher priority-fee value paid in the first tenth of block gas falls from 0.98 to 0.28. This fee compression is the identified effect. The revert-rate response is not (causally) identified and is reported descriptively. At the level of a single opportunity, the winning fee splits into an approximately invariant floor and a competitive premium. The premium scales about linearly with auction duration at high contention and less steeply at low contention. A first-price auction model with uncertain arrival and partial payments by losing searchers accounts for this pattern. We match attempts to the pool they contest. A higher-fee transaction reverts behind a lower-fee winner that landed earlier in 1.9 to 2.7% of contested opportunities by value, an upper bound. Latency therefore decides part of the ordering. A micro-auction model projects diminishing returns from shortening the window below 200 ms, as bids approach a nonzero floor. Rollups such as Arbitrum One now make auction cadence a governance tunable parameter, so these estimates inform how it is set.

Reinforcement learning (RL) is increasingly used for financial optimal-control problems when complex dynamics make analytical strategies difficult to obtain. There are financial mathematics literactures which provides many solved models whose equations and controls could evaluate and guide learning; we ask whether RL can exploit these results. We place a proximal policy optimisation (PPO) agent in an analytically solved continuous-time broker–trader game. PPO replaces the broker and chooses its trading speed while interacting with an informed trader and stochastic uninformed order flow. We derive a finite-step reward from the broker's continuous-time payoff and verify its discrete implementation through grid refinement and an exact one-step identity. With zero uninformed flow, a validation-selected PPO–FFNN approaches the reference action. With stochastic uninformed flow, the tested PPO–FFNN and PPO–LSTM remain inaccurate, although supervised learning confirms that their actors can represent the action. Monte Carlo diagnostics show that their critics do not reliably rank nearby actions; potential-based reward shaping also gives no reliable improvement. Under partial information, a causal certainty-equivalent controller based on the broker's observable history remains close to the reference, while PPO has larger errors and lower payoffs. Finally, we freeze the analytical policy and train PPO to adjust it after the execution cost changes. Halving the cost yields a repeatable improvement that closes \(2.22\%\) of the gap to the changed-cost reference. The analytical solution therefore provides both a benchmark for diagnosing RL and a useful starting policy for adaptation.

Deep reinforcement-learning policies for order execution can vary substantially across training seeds, so apparent architectural gains may reflect favourable training realisations rather than reproducible properties of the architecture. We evaluate vanilla Double Deep Q-Learning (DDQL), K-means-partitioned mixtures of DDQL experts at $K \in \{2, 4, 8\}$, and dense networks parameter-matched to the $K{=}4$ and $K{=}8$ expert budgets on 5-minute mean-aggregated BTC/USDT limit order book data from Binance. No learned configuration significantly improves mean implementation shortfall over DDQL. Under the reported specification, all have higher mean shortfall than TWAP (0.39 bps) and immediate liquidation (0.21 bps) in an environment whose frictionless replay and terminal-urgency penalty make early liquidation nearly costless; 11/100 vanilla-DDQL runs, versus none in either MoE $K{\geq}4$ arm, converge to a policy that waits until forced liquidation. We then decompose this specification on a device-matched baseline. Annealed exploration alone eliminates observed collapses (12/100 to 0/100; exact McNemar $p{=}4.9{\times}10^{-4}$), matching the elimination under expert partitioning. Combining annealed exploration with the aligned reward restores collapse in 19/30 runs; with all three specification changes, it rises to 48/100. In this environment, expert partitioning is unnecessary to suppress collapse and appears to mask a training-specification failure rather than confer an intrinsic performance benefit. No MoE $K{=}8$ run collapses under any of the six specifications tested. Across-seed dispersion is lowest at $K{=}8$ but non-monotone and not robust to family-wise adjustment, while within-policy tail risk worsens monotonically with $K$. The apparent attribution of the failure mode reverses between 30 and 100 seeds, illustrating the importance of repeated-seed evaluation.

Large language models are moving from producing trading signals to writing the code that executes them. The failure mode of the second role is silent: generated code runs, a backtest plots, yet the risk logic that the trader described is not the logic being executed. Existing code benchmarks test functional correctness on unit tests and finance benchmarks test forecasting; neither measures whether an implementation behaves like the strategy that was asked for. We introduce MintEval, a benchmark in which reference strategies are generated programmatically from a library of composable building blocks, back-translated into colloquial trader instructions, and re-implemented by the model under test. Generated and reference programs are executed bar by bar on identical market data and frictions, and compared on their actions rather than on code similarity or profit: alpha is differenced away. MintEval v0 contains 800 tasks on BTCUSDT 15-minute data, stratified by an execution-measured state-span complexity tau that is decoupled from description length. Low-cost models reach a mean ActionMatch of at most 0.544 and reproduce at most 0.087 of tasks exactly; on a stratified subset of 200 tasks a frontier model (Claude Opus 5.5) reaches 0.889 and reproduces 0.575 exactly, yet still fails silently on 0.275 of tasks. Given a menu of building blocks, models identify the strategy almost perfectly, yet 79.2% of the implementations whose specification was read correctly diverge on more than 10% of active bars. The LLM judge of a recent strategy-generation benchmark, applied verbatim, accepts every one of these silent failures.

Hybrid on-chain financial protocols are frequently evaluated with evidence that is individually useful but collectively insufficient: a unit test, transaction receipt, screenshot, service response, or several matching hashes may be presented as proof of a complete workflow even when the layers share one generated source or omit the financial authority that matters. This paper develops a manifest-bound evidence architecture for such systems. A sealed execution manifest binds declared system identities, roles, schemas, and release identity. Seven evidence layers are retained with registered derivations and linked by a common correlation identity. A conjunctive claim-promotion rule requires provenance, predicate-specific root conditions, negative controls, reproducible packaging, historical replay, and independent review in addition to hash equality. We formalize the evidence graph, prove that copied hashes cannot establish independent derivation, show how collision-resistant commitments make specified tampering detectable, define exactly-once financial-effect evidence, and distinguish original historical identities from later reproducibility attestations. The registered local case study records twelve parent Financial Interaction Assertions in one accepted cohort: 84 retained ACTUAL_EVIDENCE positions, two clean materializations with one byte-identical file index, a deterministic strict archive accepted by a separate verifier, and a selected ordered replay bound to that archive. FIA-011 is conjunctive and retains both stale-before-admission and book-epoch-change-mid-position subcases. The result demonstrates a reviewable chain of custody from financial assertion to replay-bound package while preserving the difference between observed evidence, interpretation, and future hypotheses.

A modular leveraged event-market protocol can contain individually correct contracts for risk approval, positions, debt, settlement evidence, credit pools, junior backstops, reserves, liquidation, and governance while still lacking one authoritative financial execution path. This paper develops a canonical protocol graph in which every financial domain has one storage authority and composition components coordinate transitions without recreating balances, debt, positions, evidence receipts, queues, reserves, pause state, or terminal scenario state. The architecture formalizes atomic position origination, a settlement-confirmed debt saga, trader residual release, an ordered reserve-backstop-Senior loss waterfall, loss-participating withdrawal queues, capability-conditioned execution, protocol-wide pause and timelocked recovery, deterministic deployment manifests, canonical state projection, and exactly-once reconstruction after a chain-confirmed checkpoint. We prove composition-level authority uniqueness, atomic closure, loss conservation and priority, receipt-idempotent settlement, legal no-go safety, capability-transition safety, and no-financial-reexecution recovery under explicit assumptions. The implementation is evaluated through twelve parent Financial Interaction Assertions (FIA-001-FIA-012), including mandatory FIA-011A/B capability subcases. The registered local cohort contains twelve uniquely correlated scenario records and 84 retained ACTUAL_EVIDENCE layer positions, two clean materializations with one byte-identical file index, a deterministic strict archive accepted by a separate verifier, and a selected ordered replay bound to that archive. The result connects the formal protocol graph to a reproducible evidence chain while preserving explicit boundaries between the observed engineering result and future external-operation hypotheses.

We study whether model diversity survives selection into trading. In synthetic markets with a fixed mixture of three language-model families, news presentation changes their representation among submitted orders. At the announcement round, Qwen's share of submitted orders shifts by 48 percentage points in the financing event, with little change in net order counts. In the workforce-reduction event, Mistral's share shifts by 40 percentage points while net counts reverse sign. Homogeneous populations remove opposing flow when their active decisions share a direction. An analytical decomposition shows why selection can improve or worsen price accuracy even at unchanged aggregate demand sensitivity. The evidence concerns presentation bundles and submitted flow; cleaner replication and a known-value validation are specified prospectively.

The tendency of the double auction mechanism to drive prices to competitive equilibrium has been well documented in laboratory experiments, but the phenomenon has lacked a theoretical explanation. This paper studies dynamic double auctions in a pure exchange economy where agents bid their indifference prices implied by their current holdings and preferences. We show that Walras equilibria coincide with the fixed points of the double auction and that repeated double auctions generate bounded sequences of allocations and prices whose cluster points are Walras equilibria with transfers.

Reinforcement learning for trading often struggles to balance upside participation with drawdown control. Profit-only policies can collapse toward passive long exposure on upward-drifting assets, while aggressively risk-penalized rewards can become too defensive during volatile periods. This paper proposes PPO-HRAP, a hybrid regime-aware policy that combines Proximal Policy Optimization with an interpretable regime prior. The agent observes both market features and portfolio-state variables, receives a reward combining portfolio log return, VIX-conditioned drawdown-increase penalty, target-exposure deviation, and turnover cost, and executes a blended action between the PPO actor output and a regime-derived target exposure. On the held-out 2020-2022 SPY test window, PPO-HRAP achieves 27.62% total return, 8.48% annualized return, 0.6447 Sharpe ratio, 0.8588 Sortino ratio, and 0.4592 Calmar ratio, while reducing maximum drawdown from 34.10% for Buy and Hold to 18.47%. Across five SPY seeds, PPO-HRAP remains stable with mean total return $0.2725 \pm 0.0109$ and mean Sharpe ratio $0.6219 \pm 0.0565$. Single-run cross-asset tests on QQQ and DIA further show that the proposed method ranks first on total return and Sharpe ratio for all three reported assets. These results suggest that blending learned actions with a volatility-aware regime prior is a practical way to improve risk-adjusted trading behavior, although the current policy still incurs high turnover and cross-asset robustness beyond SPY remains limited to single-run evidence.

In this paper, we extend earlier findings of supra-competitive outcomes in optimal-execution games by identifying a learned punitive mechanism that deters deviations and provides behavioural evidence of collusion. We investigate this mechanism in a two-player, finite-horizon Almgren-Chriss liquidation game. Independent proximal policy optimisation agents with access to within-episode price and action histories achieve costs below the Nash benchmark. We identify a profitable deviation by training against the mean learned liquidation schedule, then impose its first trade on one of the original agents. The opponent responds by accelerating liquidation. This response more than offsets the deviator's gain in every run and both player roles, while leaving the punisher's average payoff materially unchanged relative to not punishing under the same deviation. The punisher imposes greater losses on the deviator while preserving its own average payoff, despite the availability of more profitable, less punitive liquidation plans. Matching deviations and subsequent additional selling rise and later decline during training, while final policies retain an effective punitive response. We formalise two checks: whether punishment outweighs the gain from deviating, and whether the change in trading behaviour is large enough to account for the loss imposed. Both checks hold for the tested deviation. Together, these findings provide behavioural and economic evidence supporting a collusive interpretation of the learned supra-competitive outcomes.

Large language models are strong general-purpose code generators, but executable algorithmic trading remains a demanding specialization target: a model must translate a natural-language strategy specification into correct program logic for a specialized trading framework, execute on historical data, produce trades, and remain semantically faithful to the request. We study two complementary mechanisms for specializing language models for this setting: continued pretraining on algorithmic-trading framework code and supervised fine-tuning (SFT) on agent-validated request-to-code pairs. Evaluation is centered on QuantCode-Bench, our 400-task benchmark for Backtrader strategy generation, together with a repository-level SWE-bench-like track. Continued pretraining improves single-turn Judge Pass from 41.5% to 47.5% for Qwen3.5-397B-A17B and from 27.8% to 33.0% for Qwen3.6-35B-A3B. SFT applied after continued pretraining yields a larger gain for Qwen3.6-35B-A3B, reaching 58.2% Judge Pass and 83.5% successful backtests; in agentic evaluation it raises first-turn success from 22.3% to 58.3% and final success after up to 10 turns from 47.5% to 79.5%. Continued pretraining alone improves first-turn agentic success but lowers final success after repair from 47.5% to 32.5%, consistent with degraded instruction following, whereas SFT improves both. We also identify a capability-retention failure: domain specialization degrades parser-conformant structured tool calling, and targeted recovery SFT restores tool-call formatting but not the base checkpoint's repository-level agent performance. The results show that framework-oriented pretraining, validated SFT, and explicit capability-retention evaluation address distinct failure modes in domain-specific executable code generation.

Machine-learning signals built from financial text treat what institutions say, and what the media repeat, as evidence about value. But whoever shapes a narrative may be trading against it. We study markets with three observable voices: institutional statements (Say), media repetition (Echo) and revealed positioning (Do). We ask when words should be followed and when they should be faded. In a linear-quadratic model of an informed institution that speaks and trades before a partly credulous crowd, talking an asset down while buying it is optimal exactly when $\varphi^2<2\lambda k<\varphi$. A distribution-free identity then shows that when the observable Say-Do covariance is negative, words carry negative predictive content and should be faded. For measurement, we derive (i) an exact factorised posterior over which articles are echoes, combining arrival times with embedding similarity; (ii) a return-aligned contrastive objective that attains its bound exactly when squared embedding distances are an increasing affine function of squared outcome distances, with the tightest loss-based certificate of which neighbour rankings survive imperfect training; and (iii) a path-signature statistic for who moved first. In a controlled market with known ground truth, echo sentiment predicts returns with a significantly negative sign in all 29 simulated markets, the rolling Say-Do correlation flags false-alarm events with an AUC of 0.90, and return-aligned embeddings organise headlines by consequence rather than topic. We also report where the tools fail.

Liquidity provision in automated market makers is typically analyzed at the pool level, implicitly assuming LP homogeneity. This aggregate view can hide how liquidity provision outcomes differ between LP strategies, particularly as concentrated liquidity AMM designs operating on high-performance blockchains allow liquidity to be actively repositioned around trades. We develop a markout-based framework to decompose Uniswap LP profitability into active and passive components using two complementary methods: a LIFO subtraction method that matches short-lived mint-burn positions and attributes swap-level markouts by liquidity share; and an infinitesimal LP benchmark that estimates the performance of a fully passive, always-in-range marginal LP directly from the AMM price path. We apply these methods to Uniswap v2, v3, and v4 pools on Ethereum, Arbitrum, and Base chains, and find passive profitability can materially differ from aggregate pool profitability. In Uniswap v2, with liquidity distributed evenly and active LP behavior nearly absent, the overall and passive markouts are almost the same. In contrast, concentrated-liquidity pools have a systematic active-passive gap: passive LPs tend to underperform aggregate pool-level measures. The gap is wider on Ethereum than on L2s, consistent with active liquidity provision being more useful when block times and ordering conditions allow LPs to react to incoming flow. In general, passive LPs perform better on higher fee pools. The LIFO and infinitesimal estimates are generally consistent directionally across most pools, providing evidence of the robustness of the decomposition. The results suggest that adverse selection in AMMs is not evenly distributed among LPs, with important implications for LP strategy, fee-tier design and measurement of DEX market quality.