question about execution accuracy

[koriyyy]

Thank you for your work! I noticed that the paper did not report the execution accuracy (EX) for BIRD and Spider datasets. Could you please provide more details or clarify this?

[Satissss]

Thanks! This is an area where the experiment will improve.

Since the SPIDER and BIRD datasets only provide small-scale databases (Avg. fields<50), they lack the Schema Linking challenges found in large-scale, multi-database scenarios that LinkAlign aims to address. Thus, they are not ideal for evaluating our method. Therefore, we prioritized evaluation on the SPIDER 2.0 dataset which better reflects enterprise database environments, and we achieved SOTA performance by optimizing schema linking.

While LinkAlign outperforms traditional schema linking methods on the SPIDER and BIRD datasets, it still relies on generating SQL through the method itself, thus requireing large-scale experimental reproduction. We are refining this experiment and will provide results in the next paper version.