Forecasting Supply Chain Disruptions with Foresight Learning

TL;DR

This paper presents a new framework for training language models to forecast supply chain disruptions accurately and reliably, outperforming existing models and improving probabilistic reasoning.

Contribution

The authors introduce an end-to-end training method for LLMs to produce calibrated probabilistic forecasts of supply chain disruptions, enhancing accuracy and reasoning without explicit prompts.

Findings

Model outperforms GPT-5 in accuracy, calibration, and precision.

Training leads to more structured and reliable probabilistic reasoning.

Open-sourced evaluation dataset supports transparency.

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