Fixed typo in blog post

Author: Pseudomanifold

content/blog/rings.md

@@ -35,7 +35,7 @@ Follow along on a mathematical journey to where the grass is greener - a ~SHIRE~
 
 RINGS was forged as part of the recent paper [No Metric To Rule Them All: Towards Principled Evaluations of Graph-Learning Datasets](https://arxiv.org/abs/2502.02379) – and yes, it is a Lord of the Rings reference.
 
-Perhaps more importantly, RINGS is also a response to growing concerns in the graph-learning community about the datasets used to guide model development. Recent work has highlighted that popular graph-learning benchmarks represent a niche subset of the space of all possible graphs [(Palowitch et al., 2022)](https://doi.org/10.48550/arXiv.2203.00112), and that GNNs overfit graph structure even with it is uninformative and irrelevant to the given task [(Bechler-Speicher et al., 2024)](https://doi.org/10.48550/arXiv.2309.04332). It has also showcased instances where graph-learning methods are outperformed by those ignoring graph structure altogether [(Errica et al., 2020)](https://doi.org/10.48550/arXiv.1912.09893). These are troubling results – all the more so since they suggest a dataset environment that might not be conducive to developing and evaluating GNN models.
+Perhaps more importantly, RINGS is also a response to growing concerns in the graph-learning community about the datasets used to guide model development. Recent work has highlighted that popular graph-learning benchmarks represent a niche subset of the space of all possible graphs [(Palowitch et al., 2022)](https://doi.org/10.48550/arXiv.2203.00112), and that GNNs overfit graph structure even when it is uninformative and irrelevant to the given task [(Bechler-Speicher et al., 2024)](https://doi.org/10.48550/arXiv.2309.04332). It has also showcased instances where graph-learning methods are outperformed by those ignoring graph structure altogether [(Errica et al., 2020)](https://doi.org/10.48550/arXiv.1912.09893). These are troubling results – all the more so since they suggest a dataset environment that might not be conducive to developing and evaluating GNN models.
 
 RINGS – purpose-built from first principles – begins by establishing what an ideal dataset *is*. Namely, we assert that a good graph dataset should satisfy the following properties:
 1. The graph structure and the node features both contain task-relevant information.