TS-HINT: Enhancing Semiconductor Time Series Regression Using Attention Hints From Large Language Model Reasoning

TL;DR

TS-Hint is a novel framework that enhances semiconductor time series regression by incorporating attention hints from large language model reasoning, improving performance especially in limited data scenarios.

Contribution

The paper introduces TS-Hint, a time series foundation model with chain-of-thought reasoning that leverages attention hints to better capture temporal dynamics in semiconductor manufacturing data.

Findings

Effective in few-shot learning scenarios

Learns directly from multivariate time series features

Improves regression accuracy with limited data

Abstract

Existing data-driven methods rely on the extraction of static features from time series to approximate the material removal rate (MRR) of semiconductor manufacturing processes such as chemical mechanical polishing (CMP). However, this leads to a loss of temporal dynamics. Moreover, these methods require a large amount of data for effective training. In this paper, we propose TS-Hint, a Time Series Foundation Model (TSFM) framework, integrated with chain-of-thought reasoning which provides attention hints during training based on attention mechanism data and saliency data. Experimental results demonstrate the effectiveness of our model in limited data settings via few-shot learning and can learn directly from multivariate time series features.