Rethinking Evaluation for Temporal Link Prediction Through Counterfactual Analysis
Abstract
In response to critiques of existing evaluation methods for temporal link prediction (TLP) models, we propose a novel approach to verify if these models truly capture temporal patterns in the data. Our method involves a sanity check formulated as a counterfactual question: ``What if a TLP model is tested on a temporally distorted version of the data instead of the real data?'' Ideally, a TLP model that effectively learns temporal patterns should perform worse on temporally distorted data compared to real data. We analyse this hypothesis and introduce two temporal distortion techniques to assess six well-known TLP models.
Cite
Text
Rahman et al. "Rethinking Evaluation for Temporal Link Prediction Through Counterfactual Analysis." ICLR 2025 Workshops: ICBINB, 2025.Markdown
[Rahman et al. "Rethinking Evaluation for Temporal Link Prediction Through Counterfactual Analysis." ICLR 2025 Workshops: ICBINB, 2025.](https://mlanthology.org/iclrw/2025/rahman2025iclrw-rethinking/)BibTeX
@inproceedings{rahman2025iclrw-rethinking,
title = {{Rethinking Evaluation for Temporal Link Prediction Through Counterfactual Analysis}},
author = {Rahman, Aniq Ur and Modell, Alexander and Coon, Justin},
booktitle = {ICLR 2025 Workshops: ICBINB},
year = {2025},
url = {https://mlanthology.org/iclrw/2025/rahman2025iclrw-rethinking/}
}