The paper demonstrates that deep limit order book forecasting models can be repurposed to quantify scenario-conditioned market impact without retraining. By injecting counterfactual order‑book messages into a trained Transformer forecaster, the authors compare predictive distributions before and after the injection, defining a short‑horizon model‑implied market impact. The approach achieves a Spearman correlation of 0.99 and 97.2% directional agreement with historical outcomes for non‑neutral scenarios, and captures incremental sequence‑dependent variation beyond scenario identity and pre‑event forecasts.
Deep Limit Order Book forecasting models capture nonlinear market dynamics, but their ability to quantify the effects of counterfactual order book messages has not been systematically validated. We in...
CEDAR is a two‑stage framework for demand forecasting that incorporates planned actions and external event signals. Stage I uses an Action‑Interleaved Transformer to model controllable state transitions under interventions, while Stage II applies a Residual Correction Module that aligns event descriptions with product context using LLM‑assisted text representations. Experiments on a large Alibaba 1688 dataset show that CEDAR improves simulation accuracy over traditional time‑series forecasting baselines and benefits real‑world budget planning.
Forecast-Dojo is a replayable environment designed to benchmark and train large language model (LLM) forecasting agents. It integrates resolved prediction‑market questions with dated news, enabling agents to research events and revisit predictions at successive historical dates. The platform includes 1,568 Polymarket events, 18.8 million dated news articles, and supports repeated evaluation, training interactions, and outcome feedback, with evidence that research tools lower Brier scores across 12 tested models, though all models still lag behind historical market forecasts.
FinForesight is a diagnostic backtesting benchmark designed to evaluate deep research systems for financial forecasting. It tests four forecasting tasks—valuation, return ranking, event direction, and future financial fundamentals—across Chinese, U.S., and Hong Kong markets, while also including two diagnostic tasks for counterfactual reasoning and financial calculation. The framework limits evidence to what was available at the forecast date and compares pretraining-era, post-cutoff, and live evaluations to assess how well controlled backtests predict future performance.
WhatIf-Bench is a semi‑synthetic benchmark that tests how well time‑series forecasters use covariates to predict a target’s response to future events. It generates 37,665 tasks from mechanistic simulators of data centers, power grids, road traffic, and water networks, varying covariate numbers, historical events, and event length. Evaluation shows large‑language models outperform the best time‑series foundation models, with all models performing well on replayed events but struggling on novel covariate settings, highlighting generalization as the main challenge.
arXiv:2609.05905v1 Announce Type: cross Abstract: LLM agents are increasingly used for live forecasting, where they retrieve up-to-date information and produce estimates for unresolved future events....
The paper introduces Forecast‑Aware Residual Refinement (FARR), a method that refines long‑horizon time‑series forecasts by learning to correct predictions using endpoint evidence. FARR generates a valid forecast under normal conditions and a null forecast under a predefined null condition; the difference between these forecasts indicates how auxiliary conditioning affects the prediction. A shared residual refiner then combines this endpoint evidence with historical features to estimate the remaining error, and horizon‑wise gating controls the contribution of the predicted residual to the final output. Experiments on four long‑horizon benchmarks demonstrate that FARR achieves the best average MSE and MAE among compared methods.
Large language models use auxiliary conditioning to incorporate additional information into time-series forecasts. However, auxiliary conditioning does not consistently improve accuracy and can increa...
arXiv:2606. 18049v1 Announce Type: new Abstract: Decision-making with deep learning-based time series forecasting requires not only accurate predictions but also actionable insights.
arXiv:2512. 23847v2 Announce Type: replace-cross Abstract: We develop a statistical procedure to detect lookahead bias in economic forecasts generated by large language models (LLMs).
LEAP (Likelihood Elicitation and Aggregation for Probabilistic forecasting) is a new approach that reorganizes how evidence is used in LLM-based forecasting systems. Instead of a monolithic prediction that aggregates all evidence at once, LEAP examines each evidence item separately, elicits likelihood parameters, and combines them with an explicit prior to produce a posterior distribution. The method supports continuous, single-choice, and multi-choice forecasts and has been shown to improve prediction and calibration metrics across models on a benchmark covering forecasting, information-seeking, and browsing tasks.
The paper investigates multi‑horizon forecasting of event sequences as a conditional generative modeling problem, evaluating forecasts for both prediction accuracy and distributional realism. It introduces a benchmark with metrics, baselines, and real‑world datasets, revealing a trade‑off between accuracy and realism. The authors propose Wasserstein BaryBooster, an algorithm that combines samples from a realism‑strong diffusion model with a state‑of‑the‑art predictor to convert distributional realism into improved prediction accuracy.