We model a market with multiple dealers who compete for client order flow by dynamically updating their bid and ask quotes for a risky asset. Dealers aim to maximise expected profits while controlling inventory risk by skewing their quotes to attract offsetting order flow (internalisation) or by directly offloading positions in the market (externalisation). Using a variational approach, we derive a closed-form equilibrium for the resulting Nash competition, shedding light on key features of dealer market dynamics. We show that dealers relying on internalisation are compelled to increase their externalisation activity when competing with externalising dealers. This strategic shift in equilibrium leads to significantly higher hedging costs for all dealers and substantially wider spreads for clients.

Financial volatility exhibits substantial non-stationarity, making single-regime models inadequate for characterising changing market conditions. This paper proposes a triple-timeframe Markov-Switching GARCH (MS-GARCH) framework for volatility regime detection in EUR/USD across daily, four-hour, and hourly horizons. Three independent AR(1)-MS-GARCH models are estimated to capture macro, meso, and micro regime dynamics, while Filardo-style time-varying transition probabilities (TVTP) are incorporated at the shorter horizons through composite stress indicators. The resulting regime probabilities are combined through an outer-product construction into a 27-state cross-scale probability tensor. Using EUR/USD data from 2015-2025, the framework produces statistically distinct Calm, Turbulent, and Crisis regimes and achieves superior out-of-sample volatility forecasting performance relative to a conventional GARCH benchmark. The results suggest that volatility dynamics contain meaningful structure across multiple timescales and that modelling these scales separately provides a more informative representation of market conditions than a single-timescale approach.

Actuaries rely primarily on structured numerical data for reserving and ratemaking, while valuable predictive information in unstructured text including medical records, adjuster notes, and call transcripts remains largely unused. Manual processing of these documents is time-consuming, inconsistent across reviewers, and unscalable. We present a proof-of-concept framework using large language models (LLMs) to extract structured actuarial variables from unstructured claims data. We implement a two-stage processing architecture separating document-level extraction (Stage 1) from claim-level synthesis (Stage 2). A modular four-script Python pipeline processes synthetic FHIR-based claims data and real claims documents, extracting 36 actuarial variables across reserving, ratemaking, and claims management categories. We validate 14 core variables using two independent clinical expert reviewers scoring 20 synthetic claims on a five-point Likert rubric, achieving mean scores above 4.0 and a weighted kappa of 0.53. Integration with chain ladder reserving demonstrates practical actuarial value: severity-segmented analysis reduced reserve estimation error from 6.5% to 4.0%. The open-source implementation includes audit trails and confidence scoring, providing a replicable foundation for LLM-based actuarial variable extraction in property-casualty insurance.

We study forecasting of the realized covariation in electricity markets. The realized covariation in this context is a matrix-valued representation of the latent infinite-dimensional covariance operator and a parsimonious matrix-HAR type model is constructed to facilitate estimation. We test the model on one-week ahead forecasts of the weekly realized covariation and find that the inclusion of longer time horizons and renewable generation information adds important predictive power. We also investigate the prediction of risk premia in electricity forward markets and find that our variance forecasts provide substantially improved forecasts of spread risk premia compared to standard methods relying on backward looking volatility.

Financial decision systems require fast surrogate models for pricing, calibration, hedging, XVA, stress testing, and portfolio optimization. Standard neural surrogates reproduce prices or risk quantities, but downstream tasks depend as much on derivatives: deltas, vegas, curve and credit-spread sensitivities, exposure and objective gradients. We formulate a derivative-informed operator-learning framework in which the learned map -- a neural operator, random-feature operator, or finite-dimensional surrogate -- is trained both to match a high-fidelity pricing or risk operator and to match directional Fréchet derivatives generated on the fly. The framework combines operator learning, adjoint algorithmic differentiation, tangent sensitivity equations, random sketching of Jacobian actions, and no-arbitrage constraints. We derive error bounds showing derivative accuracy controls local stress errors, hedging error, and optimizer instability, and that discrete-time hedging error is also governed by second-order (gamma) accuracy. A Black--Scholes network over eight seeds shows a tuned derivative weight cuts vega error by 40\% and delta error by 15\% while modestly improving prices, but not an unsupervised second-order Greek. Heston and Bates random-feature experiments reduce stochastic-volatility and jump-parameter sensitivity errors by 60--76\%. A random-feature DeepONet/Galerkin operator mapping instantaneous-volatility curves to dense price surfaces reduces out-of-sample JVP error by 44\% and price RMSE by 23\% over eight seeds; it also shows derivative consistency alone does not remove no-arbitrage violations, so economic constraints must be imposed explicitly. The framework gives a disciplined route from value-only surrogates to derivative-aware engines that output differentiable instruments for hedging, risk, and control.

This paper examines the impact of market informedness on the profitability of market makers. In contrast to the existing literature, the analysis is conducted in a complex market environment featuring heterogeneous market-making agents that differ in terms of information sets and aversion to inventory risk, endogenous price formation, exogenous fundamental value dynamics, and self-exciting market order flow. The paper also establishes finite-horizon stability guarantees for the resulting state-dependent Hawkes market-taker process, including non-explosion, exponential mispricing integrability, occupation-time bounds, and a pathwise mispricing tail estimate. To address the market-making problem, the study employs a reinforcement learning framework based on the multi-agent proximal policy optimization (MAPPO) algorithm in a centralized training with decentralized execution (CTDE) setting. The study shows that informed market order flow is particularly dangerous in poorly informed markets, leading to severe adverse-selection risk. Although the complex market dynamics together with stochastic training give rise to locally non-monotonic outcomes, the results nevertheless reveal an overall upward trend in market makers' profitability as market informedness increases, suggesting that price discovery resulting from higher market informedness offsets the negative impact of adverse selection.

Per-ticker forecasting models dominate financial time-series work yet remain blind to cross-company propagation: a foundry disruption in Taiwan does not register in a single-asset model until Apple's own price has already moved. To address this limitation, we introduce a heterogeneous Rust-Python streaming architecture that maps cross-company attention as a continuous-time graph driven directly from text. We show that on the ingestion side, a zero-copy Rust edge parses news records in $\sim$100 ns and scans the target equity universe in $\sim$1.2 $μ$s. On the inference end, a multivariate Neural Hawkes Process featuring per-node continuous-time LSTM states and a bilinear latent projection propagates directed excitation, while an adaptive pruning rule bounds the computational cost of dynamic neighborhood updates. Combining these stages, we demonstrate an end-to-end processing latency of $\sim$13 ms per incoming news record on a single commodity CPU. Evaluated on a one-month temporal holdout of the FNSPID corpus (638 articles across 47 tickers), the system delivers a $1.70\times$ precision lift over random at the 90th-percentile next-day return threshold, and $3.36\times$ over a same-sector baseline. Crucially, removing the graph topology collapses precision to zero, confirming that the dynamic attention network is the sole driver of cross-company signal in this architecture.

Market stress rarely harms investors through one channel alone. Losses, volatility spikes, and deteriorating tradability often arrive together. We examine whether ESG is associated with lower exposure to clustered fragility in equity markets. Using monthly data on S&P 500 constituents from 2014 to 2025, we study downside returns, volatility, illiquidity, and a cofragility state that captures their joint occurrence within the same firm month. The evidence supports a stress-amplified resilience interpretation rather than an unconditional ESG return premium. In the return channel, the ESG association is concentrated in the extreme downside tail during stress months. In the volatility channel, higher ESG is associated with smaller risk spikes when aggregate conditions are weak. In the illiquidity channel, the association is more persistent, suggesting a liquidity-quality component whose relevance increases when market-wide trading conditions deteriorate. The central evidence comes from the joint analysis: a one-standard-deviation increase in ESG lowers the stress-period probability of severe cofragility by 0.92 percentage points, about 9% relative to the baseline. Double Machine Learning shows a similar negative ESG association after flexible adjustment for observable firm characteristics. Pillar evidence suggests stronger baseline resilience for Environmental scores and clearer stress amplification for Social scores. Overall, the findings characterize ESG as a multi-channel fragility signal for tail-risk monitoring, stress analysis, and pillar-level ESG assessment.

Bankruptcy is a low-frequency but high-impact corporate event, making early risk identification important for creditors, investors, regulators, and risk managers. Traditional bankruptcy-prediction models rely primarily on accounting ratios, but these measures may reflect financial deterioration only after it appears in reported financial statements. Narrative disclosures in annual 10-K filings may therefore provide incremental warning signals about emerging distress. This study examines whether 10-K narratives improve bankruptcy prediction beyond conventional accounting variables. Using firm-year observations matched to 10-K text, SEC financial statement data, and bankruptcy events from the Florida-UCLA-LoPucki Bankruptcy Research Database, the analysis evaluates bankruptcy risk over the year following the 10-K filing date. The paper develops a transparent Pre-Bankruptcy Stress (PB Stress) Score, a dictionary-based measure designed to capture distress-specific language related to liquidity and funding stress, debt covenant and refinancing stress, operating deterioration, restructuring and legal distress, and business fragility. The score is evaluated against a five-variable accounting baseline and a Loughran-McDonald dictionary benchmark. In the primary one-year holdout test, adding the PB Stress Score increases AUC from 0.8323 to 0.9019 and raises top-decile bankruptcy capture from 44.12% to 64.71%. The positive incremental pattern remains visible across bootstrap inference, alternative accounting benchmarks, alternative outcome definitions, and out-of-time validation. The findings indicate that distress-specific 10-K narratives provide interpretable incremental information for bankruptcy-risk monitoring beyond conventional accounting ratios.

We consider robust risk measures that arise as worst-case values of convex risk measures evaluated on uncertainty sets. We characterize continuity properties of robust risk measures through their consolidated uncertainty sets, derive dual representations for robust risk measures, and develop a set-valued dual representation for consolidated uncertainty sets. The two dual frameworks rely on distinct geometric assumptions and are therefore complementary rather than interchangeable.

Can artificial intelligence (AI) refute economic theory? I document experiments in which I asked several AI models (Gemini, Refine, Claude, and ChatGPT) to check the correctness of four published papers in economic theory, each containing an error that I helped identify or correct. ChatGPT Pro performed best, occasionally constructing counterexamples and corrected proofs, while other models fared worse. However, no model located a true error without substantial human guidance, and data contamination complicates interpretation. I argue that a competent human paired with a frontier model can outperform current peer review, but AI cannot yet refute economic theory on its own.

Large language models (LLMs) have shown strong performance across diverse financial tasks, yet portfolio management (PM), a critical financial decision-making task, remains poorly benchmarked. Existing benchmarks exhibit two main gaps: they ignore cross-asset correlation structures, thereby failing to distinguish genuinely diversified portfolios from concentrated ones, and fail to evaluate the complete PM decision pipeline in real-world scenarios. We introduce PortBench, a benchmark spanning six heterogeneous asset classes over ten years. PortBench consists of two complementary layers: a static QA dataset of 6,269 correlation-based questions across seven task templates, and a dynamic five-stage allocation pipeline that mirrors the full PM decision cycle. To evaluate these layers, we introduce two dedicated metrics: a dual-layer correlation score that measures whether proposed portfolios exploit inter-class hedging and avoid intra-class concentration, and CEPS, a metric that quantifies how reasoning errors compound across pipeline stages. We further assess strategy robustness and investor alignment under three historical stress regimes and risk profiles. Evaluating ten frontier LLMs, we find that despite strong performance on static financial QA, 90\% of model-profile combinations fail to outperform a basic equal-weight allocation, and models that satisfy every procedural constraint still suffer catastrophic drawdowns under stress. Our source code is available at \href{https://github.com/AgenticFinLab/portbench}{this https URL}.

Norway's electricity market is heavily dominated by hydropower, but the 2021-2022 energy crisis and stronger integration with Continental Europe have fundamentally altered price formation, reducing the reliability of forecasting models calibrated on historical data. Despite the critical need for updated models, a unified benchmark evaluating feature contributions across all structurally diverse Norwegian bidding zones remains lacking. Here we present a comprehensive evaluation of one-step-ahead forecasting of the Nord Pool market across all five Norwegian bidding zones. We constructed a multimodal hourly dataset spanning 2019-2025 and evaluated eight forecasting model families, including Light Gradient Boosting Machine (LightGBM), autoregressive models with exogenous variables, and advanced deep learning architectures, using a strictly causal test set. We implemented robust rolling-origin backtesting, leave-one-group-out feature ablation, and conditional regime analysis to dissect model performance and feature utility. Our results show that LightGBM achieves the best performance in every zone, with mean absolute error ranging from 1.60 to 5.58 euros per megawatt-hour, while a ridge-regularized autoregressive model with exogenous variables remains a highly competitive linear benchmark in northern zones. Feature ablation reveals that models relying solely on lagged prices and calendar variables achieve high accuracy and often match or closely approach the performance of the full multimodal model. However, conditional regime analysis demonstrates that external features like reservoir levels and gas prices remain crucial to stratify forecast errors, which consistently increase under stressed market regimes. This highlights the practical value of model interpretability and regime awareness for decision makers facing structural changes in market dynamics.

Compositional time series frequently exhibit structural breaks due to external shocks, policy changes, or market disruptions. Standard methods either ignore such breaks or handle them through fixed effects that cannot extrapolate beyond the sample, or step-function dummies that impose instantaneous adjustment. We develop a Bayesian Dirichlet ARMA model augmented with a directional-shift intervention mechanism that captures structural breaks through three interpretable parameters: a direction vector specifying which components gain or lose share, an amplitude controlling redistribution magnitude, and a logistic gate governing transition timing and speed. The model preserves compositional constraints by construction, maintains DARMA dynamics for short-run dependence, and produces coherent probabilistic forecasts through and after structural breaks. The intervention trajectory corresponds to geodesic motion on the simplex and is invariant to the choice of ILR basis. A simulation study with 400 fits across 8 scenarios shows near-zero amplitude bias and nominal 80\% credible interval coverage when the shift direction is correctly identified (77.5\% of cases); supplementary studies confirm robustness across extreme transition speeds and non-monotone DGPs. Two empirical applications to COVID-era Airbnb data characterize performance relative to simpler alternatives. Where the break is monotone and ongoing, the intervention model achieves near-nominal calibration (79.6\%) while the fixed effect substantially under-covers (66.1\%). Where post-break dynamics are non-monotone, both models are acceptably calibrated and the fixed effect outperforms on point accuracy. The intervention model's advantages are thus specific to settings with roughly monotone structural transitions.

We investigate the role of foreign direct investment (FDI) and research and development (R\&D) in the transitional dynamics of host countries using an optimal growth model. FDI may benefit the host country's GNP by enabling multinational enterprises to hire local workers. However, if the host country focuses solely on FDI, it may fall into a middle-income trap. Most importantly, we show that if the host country invests in R\&D, its economy can reach sustained growth. In this case, FDI benefits the host country, but only in the early stages of its development process.

KAN-PCA is an autoencoder that uses a KAN as encoder and a linear map as decoder. It generalizes classical PCA by replacing linear projections with learned B-spline functions on each edge. The motivation is to capture more variance than classical PCA, which becomes inefficient during market crises when the linear assumption breaks down and correlations between assets change dramatically. We prove that if the spline activations are forced to be linear, KAN-PCA yields exactly the same results as classical PCA, establishing PCA as a special case. Experiments on 20 S&P 500 stocks (2015-2024) show that KAN-PCA achieves a reconstruction R^2 of 66.57%, compared to 62.99% for classical PCA with the same 3 factors, while matching PCA out-of-sample after correcting for data leakage in the training procedure.

We introduce a general integral framework for computing fractional, complex, absolute, and logarithmic moments from the moment-generating function (MGF) under explicit regularity conditions. By evaluating a complex extension of the MGF along a vertical contour, we obtain exact integral expressions that bypass the need for explicit probability densities and high-order derivatives. We establish conditions for negative fractional moments using the symmetric Cauchy principal value, including the requirement that the distribution have no point mass at the centering point. We demonstrate the theoretical scope and computational practicality of the framework through applications to the normal-inverse Gaussian distribution and a semicontinuous compound Poisson-Gamma distribution. In the latter case, the framework handles point masses at the boundary by evaluating conditional fractional moments.

This study provides a comprehensive empirical investigation of long-range dependence (LRD) in financial markets and evaluates the ability of deep generative models to reproduce such temporal structures. Using daily data from three representative sectors--equity (S&P 500, DAX, Nikkei 225), commodities (Wheat, Corn, Soybeans), and energy (UNG, USO, XLE)--we examine the presence of LRD through three complementary approaches: rescaled range (R/S) analysis, detrended fluctuation analysis (DFA), and an ARFIMA--FIGARCH model with Student's $t$-distributed innovations. The empirical evidence suggests that while mean returns exhibit limited persistence, pronounced long memory is consistently observed in conditional volatility across most assets. Building on these findings, we assess whether Quant Generative Adversarial Networks (Quant GANs) can learn and reproduce these stylized temporal dependencies. Although the generated series successfully mimic heavy-tailed return distributions and certain aspects of volatility clustering, they generally fail to capture the magnitude and consistency of LRD observed in real data, particularly in volatility dynamics. These results highlight an important limitation of current deep generative architectures in modeling slow-decaying dependence structures and underscore the need for incorporating explicit long-memory mechanisms when synthetic financial data are intended for risk management or long-horizon forecasting applications.

This study investigates the pre-trained RNN attention models with the mainstream attention mechanisms, such as additive attention, Luong's three attentions, global self-attention and sliding window sparse attention, for the empirical asset pricing research on the top 420 large-cap US stocks. This is the first paper on the large-scale state-of-the-art (SOTA) attention mechanisms applied in the asset pricing context. They overcome the limitations of the traditional machine learning-based asset pricing, such as mis-capturing the temporal dependency and short memory. Moreover, the enforced causal masks in the attention mechanisms address the future data leaking issue ignored by the more advanced attention-based models, such as the classic Transformer. The proposed attention models also consider the temporal sparsity characteristic of asset pricing data and mitigate potential overfitting issues by deploying the simplified model structures. This provides some insights for future empirical economic research. All models are examined in three periods, which cover pre-COVID-19, COVID-19 and one year post-COVID-19, for testing the stability of these models under extreme market conditions. The study finds that in value-weighted portfolio back testing, the global self-attention model and the sliding window sparse attention model exhibit excellent capabilities in deriving the absolute returns and hedging downside risks, while they achieve an annualized Sortino ratio of 2.0 and 1.80 respectively in the period with COVID-19 in the static transaction cost scenario. Moreover, the sliding window sparse attention model performs more stably than the global self-attention model from the perspective of absolute portfolio returns with respect to the size of stocks' market capitalization.