Censoring-Aware In-Context Learning for Generalized Supplier Lead Time Estimation in Supply Chain Planning

2026-07-20

AI Summary

This study addresses the challenge of right-censored lead time data in industrial supply chains, where conventional methods discard unobserved deliveries and existing survival models exhibit limited generalizability. To overcome these limitations, this work proposes LeadTime-ICL (LT-ICL), the first framework integrating censoring awareness with in-context learning. LT-ICL leverages synthetic censored tasks for pretraining to enable zero-shot probabilistic forecasting on unseen datasets. Built upon a Transformer backbone and a conditional normalizing flow head, the model offers theoretical guarantees: its excess Continuous Ranked Probability Score (CRPS) error is bounded by prior misspecification and approximation errors. Evaluated across 24 proprietary datasets spanning seven industries, LT-ICL achieves state-of-the-art point and probabilistic prediction performance on 15 and 14 datasets, respectively, demonstrating consistently superior average rankings.

Abstract

Supplier lead time forecasting is a central input to material requirements planning, inventory optimization, and supply chain risk management. However, many industrial lead time datasets are naturally right-censored: at the time forecasts are required, some orders have not yet arrived. Standard regression and classification approaches discard this information, while conventional survival models require task-specific modeling. We propose LeadTime-ICL (LT-ICL), a censoring-aware in-context learning model for probabilistic lead time forecasting. LT-ICL combines a transformer backbone with a conditional normalizing-flow head, producing a full predictive distribution over lead times. The model is pretrained on synthetic right-censored lead time tasks, enabling in-context adaptation to new industrial datasets without task-specific parameter updates. We provide theoretical support for this formulation by showing that excess CRPS is bounded by prior misspecification and amortized approximation errors, providing clear direction for improving forecasting performance. We evaluate LT-ICL on 24 proprietary supply-chain datasets spanning seven industries. LT-ICL achieves the lowest point-forecasting error on 15 of the 24 datasets, and the lowest probabilistic forecasting error on 14 datasets, yielding the best average rank across both. These results support right-censored probabilistic forecasting as a practical formulation for supplier lead time prediction and demonstrate that pretrained in-context models can provide accurate, low-adaptation-cost forecasting for industrial planning systems.

Problem

• lead time forecasting
• right-censoring
• supply chain planning
• probabilistic forecasting
• in-context learning

Innovation

• in-context learning
• right-censored data
• probabilistic forecasting
• normalizing flow
• supply chain lead time