While we use the same data split as EMPSN, it turns out that they do not use the same data split as EGNN, although they compare themselves to EGNN. EGNN seems to use a split that was introduced in another paper and that might be widely adopted, though I will check the other papers whose results were reported in EGNN (and thus in EMPSN since the table is copied) to make sure. This investigation might reveal that we should also use the EGNN splits.
ekarais
commented
7 months ago
While we use the same data split as EMPSN, it turns out that they do not use the same data split as EGNN, although they compare themselves to EGNN. EGNN seems to use a split that was introduced in another paper and that might be widely adopted, though I will check the other papers whose results were reported in EGNN (and thus in EMPSN since the table is copied) to make sure. This investigation might reveal that we should also use the EGNN splits.