Table 3.

Additional results on the Davis and Metz dataset

Model	Davis KIKD				Metz KIKD
	RMSE	MAE	PCORR	CI	RMSE	MAE	PCORR	CI
ArkDTA (ours)	0.4979 (0.0132)	0.2863 (0.0105)	0.8176 (0.0104)	0.8684 (0.0053)	0.4127 (0.0066)	0.2559 (0.0063)	0.8336 (0.0049)	0.8430 (0.0034)
ArkDTA (Remove ${\mathcal{L}}_2$)	0.4959 (0.0071)	0.3041 (0.0072)	0.8193 (0.0061)	0.8647 (0.0034)	0.4177 (0.0099)	0.2748 (0.0084)	0.8299 (0.0085)	0.8240 (0.0079)
DeepDTA	0.4753 (0.0061)	0.2548 (0.0134)	0.8367 (0.0054)	0.8737 (0.0082)	0.4049 (0.0069)	0.2376 (0.0055)	0.8416 (0.0042)	0.8588 (0.0064)
MONN	0.5217 (0.0065)	0.3240 (0.0075)	0.7993 (0.0063)	0.8621 (0.0044)	0.4619 (0.0102)	0.3050 (0.0134)	0.7853 (0.0115)	0.8084 (0.0077)
AttentionDTA	0.4927 (0.0137)	0.2466 (0.0129)	0.8230 (0.0106)	0.8696 (0.0063)	0.3944 (0.0177)	0.2150 (0.0161)	0.8491 (0.0137)	0.8737 (0.0106)

The results are based on mean and standard deviation of its five different model’s performance on the same test instances.