← Back
Evidence source 6400Spot Checked

Towards Explainable and Reliable AI in Finance

arXiv2025-10-30Paper
Executive summary

Researchers present Time-LLM, a new time series model for financial forecasting that improves trust and transparency in AI-driven predictions. By using prompts to avoid wrong forecasts and a reliability estimator to filter out risky predictions, the system reduces costly errors. Notably, it encodes domain-specific rules for clear, auditable justifications, aiding regulatory compliance. Tests on equity and cryptocurrency markets show fewer false positives and more selective execution, though broader validation is needed for wider adoption.

What it examines

This paper explores ways to make financial forecasting with neural networks more trustworthy and understandable. It introduces methods like Time-LLM, reliability estimation, and symbolic reasoning to improve prediction accuracy and transparency, aiming to address challenges in trust and regulatory compliance in equity and cryptocurrency markets.

What it concludes

The proposed framework reduces false predictions and enables selective, reliable financial forecasts. It supports transparent and auditable AI systems, which can be used in finance for safer trading, risk management, and regulatory reporting. Future work may further improve reliability and expand applications to other financial domains.

Extracted from this source

Evidence objects

Evidence 818275% extraction confidence
Researchers unveil Time-LLM, a time series AI model for finance that uses prompts to avoid incorrect forecasts, addressing trust and transparency issues that often undermine large neural networks in financial applications.

key_findings bullet 1 · key_findings · validation V0

Evidence 818375% extraction confidence
The system uniquely combines foundation models with a reliability estimator, filtering out unreliable predictions before they reach decision-makers, and integrates symbolic reasoning to provide clear, rule-based justifications for each financial forecast.

key_findings bullet 2 · key_findings · validation V0

Evidence 818475% extraction confidence
Tested on equity and cryptocurrency markets, Time-LLM reduces false positives and enables selective execution of only the most trustworthy forecasts, though broader validation and edge case analysis are needed for wider adoption.

key_findings bullet 3 · key_findings · validation V0

Evidence 818575% extraction confidence
This paper introduces a novel time series foundation model, Time-LLM, integrating prompt-based reliability, predictive modeling, and symbolic reasoning for financial AI. Its unique architecture combines large language models, reliability estimation, and domain rule encoding, advancing explainable and selective prediction in equity and cryptocurrency forecastingmaking it compelling for transparent, regulatory-compliant financial applications.

key_findings bullet 4 · key_findings · validation V0

Raw abstract and provenance

Financial forecasting increasingly uses large neural network models, but their opacity raises challenges for trust and regulatory compliance. We present several approaches to explainable and reliable AI in finance. \emph{First}, we describe how Time-LLM, a time series foundation model, uses a prompt to avoid a wrong directional forecast. \emph{Second}, we show that combining foundation models for time series forecasting with a reliability estimator can filter our unreliable predictions. \emph{Third}, we argue for symbolic reasoning encoding domain rules for transparent justification. These approaches shift emphasize executing only forecasts that are both reliable and explainable. Experiments on equity and cryptocurrency data show that the architecture reduces false positives and supports selective execution. By integrating predictive performance with reliability estimation and rule-based reasoning, our framework advances transparent and auditable financial AI systems.

Source row: 2049 · abstract type: unknown