Can VLM Pseudo-Labels Train a Time-Series QA Model That Outperforms the VLM?
This addresses data scarcity for researchers and practitioners in time-series analysis, though it is incremental as it builds on existing VLM and noisy-label robustness methods.
The paper tackles the problem of limited labeled data in time-series question answering (TSQA) by training TSQA models with pseudo-labels generated by a vision-language model (VLM), resulting in models that outperform the VLM itself by leveraging unlabeled data.
Time-series question answering (TSQA) tasks face significant challenges due to the lack of labeled data. Alternatively, with recent advancements in large-scale models, vision-language models (VLMs) have demonstrated the potential to analyze time-series signals in a zero-shot manner. In this paper, we propose a training approach that uses pseudo labels generated by a VLM. Although VLMs can produce incorrect labels, TSQA models can still be effectively trained based on the property that deep neural networks are inherently robust to such noisy labels. Our experimental results demonstrate that TSQA models are not only successfully trained with pseudo labels, but also surpass the performance of the VLM itself by leveraging a large amount of unlabeled data.