Risk aversion of insider and dynamic asymmetric information ArXiv ID: 2512.05011 “View on arXiv” Authors: Albina Danilova, Valentin Lizhdvoy Abstract This paper studies a Kyle-Back model with a risk-averse insider possessing exponential utility and a dynamic stochastic signal about the asset’s terminal fundamental value. While the existing literature considers either risk-neutral insiders with dynamic signals or risk-averse insiders with static signals, we establish equilibrium when both features are present. Our approach imposes no restrictions on the magnitude of the risk aversion parameter, extending beyond previous work that requires sufficiently small risk aversion. We employ a weak conditioning methodology to construct a Schrödinger bridge between the insider’s signal and the asset price process, an approach that naturally accommodates stochastic signal evolution and removes risk aversion constraints. We derive necessary conditions for equilibrium, showing that the optimal insider strategy must be continuous with bounded variation. Under these conditions, we characterize the market-maker pricing rule and insider strategy that achieve equilibrium. We obtain explicit closed-form solutions for important cases including deterministic and quadratic signal volatilities, demonstrating the tractability of our framework. ...

Semantic Faithfulness and Entropy Production Measures to Tame Your LLM Demons and Manage Hallucinations ArXiv ID: 2512.05156 “View on arXiv” Authors: Igor Halperin Abstract Evaluating faithfulness of Large Language Models (LLMs) to a given task is a complex challenge. We propose two new unsupervised metrics for faithfulness evaluation using insights from information theory and thermodynamics. Our approach treats an LLM as a bipartite information engine where hidden layers act as a Maxwell demon controlling transformations of context $C $ into answer $A$ via prompt $Q$. We model Question-Context-Answer (QCA) triplets as probability distributions over shared topics. Topic transformations from $C$ to $Q$ and $A$ are modeled as transition matrices ${"\bf Q"}$ and ${"\bf A"}$ encoding the query goal and actual result, respectively. Our semantic faithfulness (SF) metric quantifies faithfulness for any given QCA triplet by the Kullback-Leibler (KL) divergence between these matrices. Both matrices are inferred simultaneously via convex optimization of this KL divergence, and the final SF metric is obtained by mapping the minimal divergence onto the unit interval [“0,1”], where higher scores indicate greater faithfulness. Furthermore, we propose a thermodynamics-based semantic entropy production (SEP) metric in answer generation, and show that high faithfulness generally implies low entropy production. The SF and SEP metrics can be used jointly or separately for LLM evaluation and hallucination control. We demonstrate our framework on LLM summarization of corporate SEC 10-K filings. ...

A High-Level Framework for Practically Model-Independent Pricing ArXiv ID: 2512.15718 “View on arXiv” Authors: Marco Airoldi Abstract We present a high-level framework that explains why, in practice, different pricing models calibrated to the same vanilla surface tend to produce similar valuations for exotic derivatives. Our approach acts as an overlay on the Monte Carlo infrastructure already used in banks, combining path reweighting with a conic optimisation layer without requiring any changes to existing code. This construction delivers narrow, practically model-independent price bands for exotics, reconciling front-office practice with the robust, model-independent ideas developed in the academic literature. ...

A Note on the Conditions for COS Convergence ArXiv ID: 2512.02745 “View on arXiv” Authors: Qinling Wang, Xiaoyu Shen, Fang Fang Abstract We study the truncation error of the COS method and give simple, verifiable conditions that guarantee convergence. In one dimension, COS is admissible when the density belongs to both L1 and L2 and has a finite weighted L2 moment of order strictly greater than one. We extend the result to multiple dimensions by requiring the moment order to exceed the dimension. These conditions enlarge the class of densities covered by previous analyses and include heavy-tailed distributions such as Student t with small degrees of freedom. ...

A Stochastic Thermodynamics Approach to Price Impact and Round-Trip Arbitrage: Theory and Empirical Implications ArXiv ID: 2512.03123 “View on arXiv” Authors: Amit Kumar Jha Abstract This paper develops a comprehensive theoretical framework that imports concepts from stochastic thermodynamics to model price impact and characterize the feasibility of round-trip arbitrage in financial markets. A trading cycle is treated as a non-equilibrium thermodynamic process, where price impact represents dissipative work and market noise plays the role of thermal fluctuations. The paper proves a Financial Second Law: under general convex impact functionals, any round-trip trading strategy yields non-positive expected profit. This structural constraint is complemented by a fluctuation theorem that bounds the probability of profitable cycles in terms of dissipated work and market volatility. The framework introduces a statistical ensemble of trading strategies governed by a Gibbs measure, leading to a free energy decomposition that connects expected cost, strategy entropy, and a market temperature parameter. The framework provides rigorous, testable inequalities linking microstructural impact to macroscopic no-arbitrage conditions, offering a novel physics-inspired perspective on market efficiency. The paper derives explicit analytical results for prototypical trading strategies and discusses empirical validation protocols. ...

Detecting AI Hallucinations in Finance: An Information-Theoretic Method Cuts Hallucination Rate by 92% ArXiv ID: 2512.03107 “View on arXiv” Authors: Mainak Singha Abstract Large language models (LLMs) produce fluent but unsupported answers - hallucinations - limiting safe deployment in high-stakes domains. We propose ECLIPSE, a framework that treats hallucination as a mismatch between a model’s semantic entropy and the capacity of available evidence. We combine entropy estimation via multi-sample clustering with a novel perplexity decomposition that measures how models use retrieved evidence. We prove that under mild conditions, the resulting entropy-capacity objective is strictly convex with a unique stable optimum. We evaluate on a controlled financial question answering dataset with GPT-3.5-turbo (n=200 balanced samples with synthetic hallucinations), where ECLIPSE achieves ROC AUC of 0.89 and average precision of 0.90, substantially outperforming a semantic entropy-only baseline (AUC 0.50). A controlled ablation with Claude-3-Haiku, which lacks token-level log probabilities, shows AUC dropping to 0.59 with coefficient magnitudes decreasing by 95% - demonstrating that ECLIPSE is a logprob-native mechanism whose effectiveness depends on calibrated token-level uncertainties. The perplexity decomposition features exhibit the largest learned coefficients, confirming that evidence utilization is central to hallucination detection. We position this work as a controlled mechanism study; broader validation across domains and naturally occurring hallucinations remains future work. ...

Hidden Order in Trades Predicts the Size of Price Moves ArXiv ID: 2512.15720 “View on arXiv” Authors: Mainak Singha Abstract Financial markets exhibit an apparent paradox: while directional price movements remain largely unpredictable–consistent with weak-form efficiency–the magnitude of price changes displays systematic structure. Here we demonstrate that real-time order-flow entropy, computed from a 15-state Markov transition matrix at second resolution, predicts the magnitude of intraday returns without providing directional information. Analysis of 38.5 million SPY trades over 36 trading days reveals that conditioning on entropy below the 5th percentile increases subsequent 5-minute absolute returns by a factor of 2.89 (t = 12.41, p < 0.0001), while directional accuracy remains at 45.0%–statistically indistinguishable from chance (p = 0.12). This decoupling arises from a fundamental symmetry: entropy is invariant under sign permutation, detecting the presence of informed trading without revealing its direction. Walk-forward validation across five non-overlapping test periods confirms out-of-sample predictability, and label-permutation placebo tests yield z = 14.4 against the null. These findings suggest that information-theoretic measures may serve as volatility state variables in market microstructure, though the limited sample (36 days, single instrument) requires extended validation. ...

Visibility-Graph Asymmetry as a Structural Indicator of Volatility Clustering ArXiv ID: 2512.02352 “View on arXiv” Authors: Michał Sikorski Abstract Volatility clustering is one of the most robust stylized facts of financial markets, yet it is typically detected using moment-based diagnostics or parametric models such as GARCH. This paper shows that clustered volatility also leaves a clear imprint on the time-reversal symmetry of horizontal visibility graphs (HVGs) constructed on absolute returns in physical time. For each time point, we compute the maximal forward and backward visibility distances, $L^{"+"}(t)$ and $L^{"-"}(t)$, and use their empirical distributions to build a visibility-asymmetry fingerprint comprising the Kolmogorov–Smirnov distance, variance difference, entropy difference, and a ratio of extreme visibility spans. In a Monte Carlo study, these HVG asymmetry features sharply separate volatility-clustered GARCH(1,1) dynamics from i.i.d.\ Gaussian noise and from randomly shuffled GARCH series that preserve the marginal distribution but destroy temporal dependence; a simple linear classifier based on the fingerprint achieves about 90% in-sample accuracy. Applying the method to daily S&P500 data reveals a pronounced forward–backward imbalance, including a variance difference $Δ\mathrm{“Var”}$ that exceeds the simulated GARCH values by two orders of magnitude and vanishes after shuffling. Overall, the visibility-graph asymmetry fingerprint emerges as a simple, model-free, and geometrically interpretable indicator of volatility clustering and time irreversibility in financial time series. ...

AI-Trader: Benchmarking Autonomous Agents in Real-Time Financial Markets ArXiv ID: 2512.10971 “View on arXiv” Authors: Tianyu Fan, Yuhao Yang, Yangqin Jiang, Yifei Zhang, Yuxuan Chen, Chao Huang Abstract Large Language Models (LLMs) have demonstrated remarkable potential as autonomous agents, approaching human-expert performance through advanced reasoning and tool orchestration. However, decision-making in fully dynamic and live environments remains highly challenging, requiring real-time information integration and adaptive responses. While existing efforts have explored live evaluation mechanisms in structured tasks, a critical gap remains in systematic benchmarking for real-world applications, particularly in finance where stringent requirements exist for live strategic responsiveness. To address this gap, we introduce AI-Trader, the first fully-automated, live, and data-uncontaminated evaluation benchmark for LLM agents in financial decision-making. AI-Trader spans three major financial markets: U.S. stocks, A-shares, and cryptocurrencies, with multiple trading granularities to simulate live financial environments. Our benchmark implements a revolutionary fully autonomous minimal information paradigm where agents receive only essential context and must independently search, verify, and synthesize live market information without human intervention. We evaluate six mainstream LLMs across three markets and multiple trading frequencies. Our analysis reveals striking findings: general intelligence does not automatically translate to effective trading capability, with most agents exhibiting poor returns and weak risk management. We demonstrate that risk control capability determines cross-market robustness, and that AI trading strategies achieve excess returns more readily in highly liquid markets than policy-driven environments. These findings expose critical limitations in current autonomous agents and provide clear directions for future improvements. The code and evaluation data are open-sourced to foster community research: https://github.com/HKUDS/AI-Trader. ...

Does it take two to tango: Interaction between Credit Default Swaps and National Stock Indices ArXiv ID: 2512.07887 “View on arXiv” Authors: Yhlas Sovbetov, Hami Saka Abstract This paper investigates both short and long-run interaction between BIST-100 index and CDS prices over January 2008 to May 2015 using ARDL technique. The paper documents several findings. First, ARDL analysis shows that 1 TL increase in CDS shrinks BIST-100 index by 22.5 TL in short-run and 85.5 TL in long-run. Second, 1000 TL increase in BIST index price causes 25 TL and 44 TL reducation in Turkey’s CDS prices in short- and long-run respectively. Third, a percentage increase in interest rate shrinks BIST index by 359 TL and a percentage increase in inflation rate scales CDS prices up to 13.34 TL both in long-run. In case of short-run, these impacts are limited with 231 TL and 5.73 TL respectively. Fourth, a kurush increase in TL/USD exchange rate leads 24.5 TL (short-run) and 78 TL (long-run) reductions in BIST, while it augments CDS prices by 2.5 TL (short-run) and 3 TL (long-run) respectively. Fifth, each negative political events decreases BIST by 237 TL in short-run and 538 TL in long-run, while it increases CDS prices by 33 TL in short-run and 89 TL in long-run. These findings imply the highly dollar indebted capital structure of Turkish firms, and overly sensitivity of financial markets to the uncertainties in political sphere. Finally, the paper provides evidence for that BIST and CDS with control variables drift too far apart, and converge to a long-run equilibrium at a moderate monthly speed. ...