arXane › Quantitative Finance › Quantitative Finance
Wednesday, 4 November 2026: 3 new papers, 6 cross-lists and 0 replacements in Statistical Finance (q-fin.ST), in arXiv's order, with full abstracts. The same papers as arXiv's own listing, set as a newspaper; every title links to arXiv.
We develop PanelGP, a conditional factor model for large panels of stock returns in which firm characteristics determine the factor loadings and economic conditions determine the expected factors. Both relations are estimated nonparametrically with Gaussian-process priors. The kernel choice nests instrumented principal component analysis (IPCA) as well as smooth nonlinear maps. Asset-specific loading innovations and common factor innovations separate idiosyncratic from systematic risk, so one estimated model forecasts both expected returns and their covariance matrix. Estimation uses sparse variational inference with a joint posterior over loadings and factors. Because each predicted return is bilinear in a Gaussian state, the predictive covariance matrix and the cumulants of any portfolio have closed forms. In simulations, the estimator recovers the nonlinear maps and the variance parameters. For S&P 500 constituents over 1995-2024, nonlinear loadings rank stocks better than affine loadings and improve the Gaussian log score. Relative to IPCA, the resulting portfolios have higher Sharpe ratios, also net of trading costs, and the minimum-variance portfolio has lower volatility. The predicted variances rank stocks and months by risk but are too compressed.
Researchers ask large language model (LLM) research agents whether a forecasting model beats a static benchmark, a verdict that real data cannot check. We treat the agent as a statistical test and score pre-specified runs of three frontier models and one open-weight model in simulated worlds with a planted signal. The agents detect weak signals more often than a pre-specified out-of-sample test, mainly because they accept weaker evidence: held to an agent's false-positive rate, a conventional in-sample test does as well or better. The agents claim predictability too often in worlds without a signal, state miscalibrated probabilities, and change their verdicts after a sentence about prior research; the magnitudes depend on the model. The excess false claims persist in simulated worlds without a resampling artifact. Agent verdicts are informative but need a neutral brief, a decision rule fixed outside the agent and a sealed holdout.
We quantify the response of market volatility to a preceding price trend as a function of lag at minute resolution. The asymmetry of the response is investigated. The sample consists of trades in three of the largest cryptocurrencies on a single exchange and in three large US stocks and the S&P 500 exchange-traded fund between 2020 and 2026. For the cryptocurrencies the asymmetry is positive, matching the sign of the leverage effect. It follows the trend: it is greatest at the first measurable lag, 20 minutes, and weak at negative lags, where the trend follows the volatility. At those lags a scan over window lengths resolves a small component of opposite sign that has not yet been disentangled from what the statistic itself generates. For Bitcoin and Ether the asymmetry decays to one half by a lag of roughly an hour and a half. Decomposition under time reversal divides the asymmetry into a T-even part and a T-odd part. Only the latter carries an arrow of time. For Bitcoin and Ether the T-odd part is detected. At the first lag it is between 0.4 and 0.9 of the T-even part at windows of 10 and 20 minutes. At 20 minutes this is the ratio produced by a purely one-sided response in this statistic: at that window the response is consistent with being wholly one-sided in time. None of the equities exhibits a detectable T-odd part at 20 minutes in the straight-line fit over 2.5 trend scales. This fit is the shared instrument of comparison. The fund's return-variance correlation displays an arrow of time; the cryptocurrencies' asymmetry is independent of the shapes of the fits, and Bitcoin's and Ether's of their ranges. On simulated and shuffled data without coupling, the analysis yields zero with errors of the correct size. It also recovers the time scale of a one-sided response embedded in a simulated market.
Financial workflows need fast, calibrated, auditable decisions, not just generated text. We train System-1 decision models, small language models fine-tuned to answer typed questions (choice/score/noul) with a probability in a single forward pass, across two model families (K2-Horizon and Qwen3.5/3.8) and nine rungs from 0.8B to 32B parameters, using one frozen LoRA recipe on a public decision benchmark and two new finance corpora: a programmatic-gold policy corpus (rule-reading with exact labels) and a reality-labeled outcome corpus (19,640 realized forward-return questions in walk-forward splits, plus a sealed 2026 test of 1,360 questions). Narrow fine-tuning costs general decision competence 9-45 percentage points, a data-mixture effect that shrinks with model scale and is associated with narrow-gradient volume. On reality-labeled questions, every trained rung from 0.8B to 32B lands at the horizon-conditioned prior baseline (accuracy 0.37-0.38, no detectable conditional edge under our controls); the models respond with flat, well-calibrated probabilities that track per-horizon base rates: naturally honest uncertainty, with a mid-scale calibration sweet spot (ECE 0.026 at 2B, degrading to 0.043-0.044 at 9-32B under prior drift). A mixed-corpus remediation carries all competencies in one adapter at modest cost: it matches or slightly exceeds the generalist specialist on the public benchmark while holding the policy ceiling and near-specialist outcome calibration, yielding a deployable artifact with per-domain temperature regimes and a band-dispatch escalation analysis: sharp, dispatchable bands where rules determine answers; flat, review-routing bands where reality does not.
Quantum reservoir computing (QRC) processes temporal information through quantum dynamics, but it remains unclear whether small QRC models can forecast real data competitively with classical models and whether their predictions survive the transfer from ideal simulation to noisy hardware. We address both questions for a six-qubit gate-based reservoir that forecasts next-day motor-insurance claim counts. Five reservoir candidates were selected and frozen using hardware-native interaction graphs, Trotterized Ising dynamics, memory and measurement-robustness criteria, circuit cost, and calibrated-noise screening. On 232 held-out test days, the ideal candidates reached root-mean-square errors (RMSEs) of 5.66-6.12 claims/day, compared with 5.36 for ridge regression. For four of the five, the difference from ridge was not significant under a 95% block bootstrap, and their accuracy was comparable with classical recurrent and reservoir baselines. On the 156-qubit IBM Kingston processor, we ran 63 evaluations in 21 paired rounds, comparing hardware-aware, random-native, and Qiskit optimization level 1 layouts. Hardware-aware placement gave lower RMSE than Qiskit in every round, with a maximum transfer distortion of 7.59 claims/day (11.84 for Qiskit), and avoided the severe transfer failures seen for some random layouts. Most of the hardware distortion appeared as a systematic forecast shift, which a short hardware probe on training inputs detected before the test runs. Longer replay windows raised circuit cost without a consistent gain in forecasting accuracy or hardware transfer. Small QRC models can thus reach competitive accuracy on this task, and hardware-aware layout selection makes their transfer to IBM Kingston more consistent.
Large language models (LLMs) read company news at scale, and backtests suggest that their readings predict stock returns. We ask how much of that predictability survives a test that rules out look-ahead, memorization and selection. We propose OLIVIA, a protocol for such tests: point-in-time inputs with company identities masked, selection on validation data only, an analysis plan frozen by cryptographic hash, and a test period unblinded once through a logged lock. We apply it to AI decision models, non-generative language models that return probabilities for typed questions, reading news about S&P 500 stocks: an open-weights model we fine-tuned on 2024-2025 data, three commercial models and standard baselines. The fine-tuned model's signal replicated on two validation quarters (daily information coefficients 0.055 and 0.065) but fell to 0.020 (t = 1.7) on the blinded January-July 2026 test. The commercial models' readings line up with the overnight move to the open (information coefficient 0.074 to 0.077) but not with the move after it, so the news is priced by the time a trader can act on it at the open. No model can tell over- from under-reaction, and no pre-registered strategy survives costs and a deflated Sharpe ratio. Commercial models identify masked companies in 62% to 74% of cases against a 28% baseline: masking hides names, not companies. The market reads the news first.
Under temporal nonstationarity, model performance rankings reverse across training horizons due to heterogeneous sensitivities to memory. While existing literature addresses adaptive lookbacks, effective model complexity, and empirical forecast evaluation in isolation, no unified framework explains the structural mechanics driving these rank reversals. We bridge this gap by deriving a model-agnostic decomposition of out-of-sample $R^2$ into target alignment, miscalibrated aggressiveness, and predictive instability. Within this framework, we mathematically characterize complexity phase boundaries, establish closed-form optimal-memory equations for ridge regression under local drift, and extend these guarantees to complex model architectures via surrogate bounds. We validate these theoretical dynamics by benchmarking 14 linear, tree-based, and neural architectures across large-scale financial panels. In doing so, our framework demonstrates that empirical model rank reversals are driven by predictable surges in miscalibrated aggressiveness, providing a rigorous mechanistic diagnosis of generalization under temporal shift.
Sequential forecast intervals are often evaluated by aggregate coverage even when calibration memory is bounded. This can be misleading when economically important states recur after long absences: a recent-only buffer may satisfy its overall coverage target while no longer containing errors from earlier episodes of the returning state. We study this problem in next-day financial range forecasting, holding the forecasting model, conformity score, adaptive update, and memory budget constant while varying the historical residual information available at each query. Recent-only first-in, first-out (FIFO) calibration is near nominal in aggregate but undercovers substantially during high-volatility states in both an exchange-traded fund development panel and a fixed stock panel. An unbounded state-history reference yields a lower observed stress-period interval score; Rare replay yields lower high-VIX interval scores than the matched Recent-n and Hybrid-n controls in both panels, and the state-structured versus recent-only contrast is largest near stress re-entry. Bounded calibration memory is therefore part of the forecasting information design, and aggregate calibration alone can conceal state-specific reliability failures.
Conventional public order-book data summarize trading without identifying the accounts behind it. Hyperliquid's Level-4 record also exposes persistent pseudonymous wallet addresses. We reconstruct full-depth books from consensus-ordered messages and evaluate wallet identity as a feature for short-horizon return prediction. The July 1-27, 2026 record covers 17.1 billion messages, 14.3 million aggressive orders by 147,113 wallets, and $84.3 billion in taker notional across ten markets. In the three model markets, notional-weighted ten-second markout rankings persist across adjacent ten-day windows (Spearman rho = 0.52); covariate-adjusted validation markouts reach 3.11 basis points in the top ventile. On a 100-millisecond grid, eleven features describing previously scored top-decile wallets raise one-second out-of-sample R2 from 10.88% to 12.