Paper Detail
Benjamin Turtel, Paul Wilczewski, Kris Skotheim
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
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@misc{turtel2026forecasting,
title = {Forecasting Supply Chain Disruptions with Foresight Learning},
author = {Benjamin Turtel and Paul Wilczewski and Kris Skotheim},
year = {2026},
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 substanti},
url = {https://huggingface.co/papers/2604.01298},
keywords = {large language models, probabilistic forecasts, calibration, supply chain disruptions, domain-specific adaptation, decision-ready signals, huggingface daily},
eprint = {2604.01298},
archiveprefix = {arXiv},
}
{}