In Task 2, when determining the final rankings, are the scores from various metrics (CIDEr, SPICE, METEOR, ROGUE, BLEU-4, BLEU-3, BLEU-2, and BLEU-1) combined using a weighted sum?
Or are the rankings primarily based on the CIDEr scores, with ties resolved by sequentially considering the scores from the other metrics in a predetermined order of priority?
Thank you.
Hi,
The scores will be considered sequentially in the order of priority.
Basically, if there is no tied entry based on CIDEr score, the ranking will be determined according to CIDEr score, as shown on the leaderboard.
The same method of determining rankings applies to both topics.
Thanks.