This paper examines the impact of introducing a Rank-Dependent Utility (RDU) agent into a von Neumann-Morgenstern (vNM) pure-exchange economy with no aggregate uncertainty. In the absence of the RDU agent, the classical theory predicts that Pareto-optimal allocations are full-insurance, or no-betting, allocations. We show how the probability weighting function of the RDU agent, seen as a proxy for probabilistic risk aversion that is not captured by marginal utility of wealth, can lead to Pareto optima characterized by endogenous betting, despite common baseline beliefs. Such endogenous betting at an optimum leads to uncertainty-generating trade arising purely from heterogeneity in the perception of risk, rather than in beliefs. Our results formalize the intuitive understanding that probability weighting can act as an endogenous source of belief heterogeneity, and provide a new behavioral foundation for the coexistence of common beliefs and speculative behavior, in an environment with no initial aggregate uncertainty. Interpreting the RDU agent's nonlinear weighting function as an ``internality'' prompts the question of whether a social planner should intervene. We show how a benevolent social planner can nudge the RDU agent to behave closer to a vNM agent, through costly statistical or financial education, thereby (partially) restoring the optimality of full-insurance allocations.

Foundation models have transformed domains from language to genomics by learning general-purpose representations from large-scale, heterogeneous data. We introduce TradeFM, a 524M-parameter generative Transformer that brings this paradigm to market microstructure, learning directly from billions of trade events across >9K equities. To enable cross-asset generalization, we develop scale-invariant features and a universal tokenization scheme that map the heterogeneous, multi-modal event stream of order flow into a unified discrete sequence -- eliminating asset-specific calibration. Integrated with a deterministic market simulator, TradeFM-generated rollouts reproduce key stylized facts of financial returns, including heavy tails, volatility clustering, and absence of return autocorrelation. Quantitatively, TradeFM achieves 2-3x lower distributional error than Compound Hawkes baselines and generalizes zero-shot to geographically out-of-distribution APAC markets with moderate perplexity degradation. Together, these results suggest that scale-invariant trade representations capture transferable structure in market microstructure, opening a path toward synthetic data generation, stress testing, and learning-based trading agents.

We document the first systematic evidence of negative spillover effects in crypto asset returns across blockchains. Using on-chain data from Ethereum, Solana, Binance Smart Chain, Arbitrum, and Avalanche (2022-2025), we show that surges on one chain often coincide with declines on others, in contrast to the positive co-movements typical of equity markets. These spillovers intensify during attention shocks, proxied by chain activity and extreme return events, and persist after controlling for global equity returns, interest rates, and Bitcoin. Nonlinear factor models reveal that attention-driven capital reallocation, rather than common information, underlies these dynamics. Our findings introduce a new form of cross-market linkage, attention-induced substitution, that shapes risk transmission in crypto markets. The results carry implications for portfolio diversification, systemic risk measurement, and regulation of token launches that may trigger cross-chain capital flight.

Mention markets, a type of prediction market in which contracts resolve based on whether a specified keyword is mentioned during a future public event, require accurate probabilistic forecasts of keyword-mention outcomes. While recent work shows that large language models (LLMs) can generate forecasts competitive with human forecasters, it remains unclear how input context should be designed to support accurate prediction. In this paper, we study this question through experiments on earnings-call mention markets, which require forecasting whether a company will mention a specified keyword during its upcoming call. We run controlled comparisons varying (i) which contextual information is provided (news and/or prior earnings-call transcripts) and (ii) how \textit{market probability}, (i.e., prediction market contract price) is used. We introduce Market-Conditioned Prompting (MCP), which explicitly treats the market-implied probability as a prior and instructs the LLM to update this prior using textual evidence, rather than re-predicting the base rate from scratch. In our experiments, we find three insights: (1) richer context consistently improves forecasting performance; (2) market-conditioned prompting (MCP), which treats the market probability as a prior and updates it using textual evidence, yields better-calibrated forecasts; and (3) a mixture of the market probability and MCP (MixMCP) outperforms the market baseline. By dampening the LLM's posterior update with the market prior, MixMCP yields more robust predictions than either the market or the LLM alone.

Prediction markets provide a unique setting where event-level time series are directly tied to natural-language descriptions, yet discovering robust lead-lag relationships remains challenging due to spurious statistical correlations. We propose a hybrid two-stage causal screener to address this challenge: (i) a statistical stage that uses Granger causality to identify candidate leader-follower pairs from market-implied probability time series, and (ii) an LLM-based semantic stage that re-ranks these candidates by assessing whether the proposed direction admits a plausible economic transmission mechanism based on event descriptions. Because causal ground truth is unobserved, we evaluate the ranked pairs using a fixed, signal-triggered trading protocol that maps relationship quality into realized profit and loss (PnL). On Kalshi Economics markets, our hybrid approach consistently outperforms the statistical baseline. Across rolling evaluations, the win rate increases from 51.4% to 54.5%. Crucially, the average magnitude of losing trades decreases substantially from 649 USD to 347 USD. This reduction is driven by the LLM's ability to filter out statistically fragile links that are prone to large losses, rather than relying on rare gains. These improvements remain stable across different trading configurations, indicating that the gains are not driven by specific parameter choices. Overall, the results suggest that LLMs function as semantic risk managers on top of statistical discovery, prioritizing lead-lag relationships that generalize under changing market conditions.

Contemporary AI regulation, including the EU Artificial Intelligence Act and related governance frameworks, increasingly requires institutions to justify the training data used in automated decision-making. Yet existing governance regimes provide limited operational methods for selecting, weighting, and explaining data inputs. We introduce the Smart Data Portfolio (SDP) framework, which treats data categories as productive but risk-bearing assets, formalizing input governance as an information-risk trade-off. Within this framework, we define two portfolio-level quantities, Informational Return and Governance-Adjusted Risk, whose interaction characterizes attainable data mixtures and yields a Governance-Efficient Frontier. Regulators shape this frontier through risk caps, admissible categories, and weight bands that translate fairness, privacy, robustness, and provenance requirements into measurable constraints on data allocation while preserving model flexibility. A sectoral illustration shows how different AI services require distinct portfolios within a common governance structure. The framework provides an input-level explanation layer through which institutions can justify governed data use in large-scale AI deployment.

We present a generative framework for pricing European-style basket options by learning the conditional terminal distribution of the log arithmetic-weighted basket return. A Mixture Density Network (MDN) maps time-varying market inputs encoded via truncated path signatures to the full terminal density in a single forward pass. Traditional approaches either impose restrictive assumptions or require costly re-simulation whenever inputs change, limiting real-time use. Trained on Monte Carlo (MC) under GBM with time-varying volatility or local volatility, the MDN acts as a reusable surrogate distribution: once trained, it prices new scenarios by integrating the learned density. Across maturities, correlations, and basket weights, the learned densities closely match MC (low KL) and produce small pricing errors, while enabling \emph{train-once, price-anywhere} reuse at inference-time latency.

Large language models (LLMs) excel at generating human-like responses but often struggle with interactive tasks that require access to real-time information. This limitation poses challenges in finance, where models must access up-to-date information, such as recent news or price movements, to support decision-making. To address this, we introduce Financial Agent, a knowledge-grounding approach for LLMs to handle financial queries using real-time text and tabular data. Our contributions are threefold: First, we develop a Financial Context Dataset of over 50,000 financial queries paired with the required context. Second, we develop FinBloom 7B, a custom 7 billion parameter LLM, by fine-tuning Bloom 7B on 14 million financial news articles from Reuters and Deutsche Presse-Agentur (DPA), alongside a random sample of 25% from 12 million Securities and Exchange Commission (SEC) filings. Third, we fine-tune FinBloom 7B using the Financial Context Dataset to serve as a Financial Agent. This agent generates relevant financial context, enabling efficient real-time data retrieval to answer user queries. By reducing latency and eliminating the need for users to manually provide accurate data, our approach significantly enhances the capability of LLMs to handle dynamic financial tasks. Our proposed approach makes real-time financial decisions, algorithmic trading and other related tasks streamlined, and is valuable in contexts with high-velocity data flows.

Neural operator (NO) architectures learn nonlinear maps between infinite-dimensional function spaces and are widely used to accelerate simulation and enable data-driven model discovery. While universality results ensure expressivity, they do not address \emph{complexity}: for broad operator classes described only through regularity (e.g.\ uniform continuity or $C^r$-regularity), information-theoretic lower bounds imply that minimax-optimal NO approximation rates scale \emph{exponentially} in the reciprocal accuracy $1/\varepsilon$. This has shifted the focus of NO theory toward identifying additional problem-specific structure, beyond regularity, under which suitably tailored NO architectures can leverage to unlock polynomial scaling in $1/\varepsilon$. We exhibit the first polynomial-scaling regime for NO approximations of solution operators in stochastic analysis; by identifying structured families of \emph{non-Markovian} BSDEs with randomized terminal condition parameterized by the Sobolev-regular terminal condition and by Sobolev-regular additive nonlinear perturbations of the generator. We prove that their solution operator can be approximated (uniformly over the family) by a tailored NO whose number of trainable parameters grows \emph{polynomially} in $1/\varepsilon$. We unlock this polynomial scaling regime by \emph{informing the NO's inductive bias} by factoring out the singular part of the associated semilinear elliptic PDE Green's function and by incorporating the Doléans--Dade exponential of the BSDE's common non-Markovian factor into the NO's decoding layers. As a byproduct, we extend polynomial-scaling guarantees from families of linear elliptic PDEs on regular domains to the semilinear setting.

To comply with increasingly stringent international standards in risk management and regulation, several approaches have been developed in the literature for forecasting tail-risk measures such as Value-at-Risk (VaR) and Expected Shortfall (ES). However, the accuracy of these measures can be significantly affected by multiple sources of uncertainty, including model misspecification, data limitations and estimation procedures. To address these challenges and enhance the predictive performance of individual models, this study introduces novel forecast combination strategies based on the Model Confidence Set (MCS) methodology. Specifically, a strictly consistent joint VaR-ES loss function is employed to identify the best-performing models, which constitute the Set of Superior Models (SSM). Subsequently, the VaR and ES forecasts of the models included in the SSM are combined using various weighting schemes. An empirical analysis based on nine stock market indices at the 2.5\% and 1\% risk levels provides evidence that the proposed combined predictors are a robust alternative for forecasting tail-risk measures, successfully passing standard backtests and consistently entering the SSM of the MCS.

We construct a stochastic dynamical systems theory in which sustainability is a structural boundary property of a fully coupled Earth--Human--Production system. Each subsystem is modelled as a vector-valued process governed by stochastic differential equations with multiplicative noise and absolute bidirectional cross-subsystem flows. Biodiversity is endogenous, and societal evaluation is represented by a reflexive functional whose weights depend on evolving human capabilities. Sustainability, development, and sustainable development are defined as trajectory properties. Sustainability corresponds to boundary non-attainment with positive or unit probability; development corresponds to local ascent in the evaluation functional; sustainable development requires directional alignment under strictly positive survival probability. No optimisation problem is imposed. Necessary and sufficient conditions are derived using Feller boundary classification and stochastic Lyapunov methods. A central result identifies the sign of the net absolute cross-subsystem flow on each component as a phase-transition parameter: if negative near zero, the boundary is of exit type and almost-sure persistence is structurally impossible, independent of intrinsic regeneration, capability accumulation, or productivity parameters. Because flows are absolute, any perturbation diffuses through the entire coupled system without requiring correlated exogenous shocks. The reflexive evaluation structure generically induces non-transitive development relations, providing a formal mechanism for path-dependent welfare comparisons. Sustainability emerges as a geometric property of boundary structure and vector-field alignment, not as a corollary of intertemporal optimality.

Climate-driven flood risk increasingly necessitates managed retreat through government buyout programmes, yet empirical evidence documents substantial racial and economic disparities in programme implementation. Here we develop a three-level Stackelberg game to analyse how federal-local cost-sharing arrangements generate inequitable outcomes through strategic interactions among federal authorities, local governments, and heterogeneous homeowners. Our model reveals three distinct mechanisms driving inequity: differential discount rates across income groups, local governments' tax-base preservation incentives, and participation thresholds that exclude fiscally constrained communities. Numerical analysis of 34,493 households across nine flood-prone US regions demonstrates that the current Federal Emergency Management Agency 75/25 cost-sharing arrangement produces a relocation ratio gap of 0.26--low-income households relocate at roughly one-quarter the rate of high-income households. Achieving near-equity requires federal cost shares of at least 85%, though equity-weighted mechanisms can attain similar outcomes at 25% lower cost. These findings provide a theoretical foundation for understanding observed disparities and identify policy levers for more equitable climate adaptation.

We study how increases in remote work opportunities for men affect their spouses' labor supply. Exploiting variation in the change in work-from-home (WFH) exposure across occupations before and after the COVID-19 pandemic, we find that increases in men's WFH exposure led to sizable improvements in their wives' labor-market outcomes: annual employment rose by roughly 2.5 percentage points (from a 69% pre-treatment mean), earnings increased by about 5%, weekly hours worked rose by roughly half an hour, weeks worked increased by about 1.3%, and the likelihood of part-time work declined by approximately 9%. Evidence from time-use diaries and childcare questionnaires suggests these effects are driven by intra-household reallocation of child-caring time: women are less likely to engage in primary childcare activities, while men working at home partially compensate by covering more for their spouse. These results highlight the role of households in shaping the labor market consequences of remote work.

How does transport cost affect the spatial organization of economic activities? This study develops a theoretical framework that distinguishes between two types of dispersion forces in spatial models: "local" dispersion forces acting within cities, and "global" dispersion forces acting across them. The distinction leads to a systematic classification of spatial models into a few fundamental types, each with distinct endogenous spatial patterns and comparative statics in response to changes in transport costs. The framework reconciles empirical findings and clarifies how transport-induced reorganization of economic activities can depend on the spatial scale of dominant dispersion forces.