Forecasting Supply Chain Disruptions with Foresight Learning

2026-04-01

AI Summary

This work addresses the challenge of predicting rare yet high-impact supply chain disruption events from noisy, unstructured data. The authors propose an end-to-end framework that trains large language models using real-world disruption outcomes as supervision, thereby inducing structured probabilistic reasoning capabilities without explicit prompting. By integrating supervised learning with probabilistic calibration, the method significantly enhances prediction accuracy, calibration, and precision. Empirical results demonstrate superior performance over strong baselines, including GPT-5. To promote research transparency and reproducibility, the authors also release their evaluation dataset publicly.

Abstract

Anticipating supply chain disruptions before they materialize is a core challenge for firms and policymakers alike. A key difficulty is learning to reason reliably about infrequent, high-impact events from noisy and unstructured inputs - a setting where general-purpose models struggle without task-specific adaptation. We introduce an end-to-end framework that trains LLMs to produce calibrated probabilistic forecasts using realized disruption outcomes as supervision. The resulting model substantially outperforms strong baselines - including GPT-5 - on accuracy, calibration, and precision. We also show that training induces more structured and reliable probabilistic reasoning without explicit prompting. These results suggest a general pathway for training domain-specific forecasting models that produce decision-ready signals. To support transparency we open-source the evaluation dataset used in this study. Dataset: https://huggingface.co/datasets/LightningRodLabs/supply-chain-predictions

Problem

• supply chain disruptions
• forecasting
• rare events
• probabilistic reasoning
• noisy inputs

Innovation

• foresight learning
• probabilistic forecasting
• supply chain disruption
• large language models
• calibrated predictions