FS-MOL result plot different from paper

[invariant0]

Hi!

I would like to confirm whether the plot for in-domain fs-mol datasets on the paper are coming from the test_results folder of this repo? I run the plot script and I think the results are little bit lower than the plots on the paper?

Thanks!

[invariant0]

To make it more clear, I will present screenshot for illustration

I think this is on the paper and below is the plot comming from the script, for example the best performance for 32 few shot is above 0.3 in the paper it appears and below 0.3 in the plot using test_results folder

Thanks again for your reply

[BFeng14]

Hi, r2 and R2os is difined in different way, where r2 = max(0, pearson_corr)**2, you can plot the r2 result and check it~

[invariant0]

Thanks very much for pointing out my oversight, however I happen to found your metric defining function and it seems you are using pearson correlation to calculate r2 instead of the one you mention. Thanks for your clarification.

[BFeng14]

sorry for the previous typo, and this r2 is also defined in the paper, thank you for your correction