[Rspamd-Users] Higher Neural Scores in 2.5

[Tim Harman]

Hi,

I've noticed since the upgrade to 2.5 I'm getting (what I consider) 
really high scores from Neural:

NEURAL_HAM_SHORT (-5.101642) [-2.551]
BAYES_HAM (-2.99843) [99.99%]
ARC_ALLOW (-1) [lists.rspamd.com:s=arc:i=1]

NEURAL_HAM_SHORT (-5.31841) [-2.659]
BAYES_HAM (-2.994811) [99.97%]
ARC_REJECT (2) [signature check failed: fail, {[1] = 
sig:google.com:reject}]

I've since removed the weight of "2/-2" for NEURAL_HAM/SPAM_SHORT to 
account for this.

Does anyone else notice that 2.5 gives higher scores to NEURAL?

Thanks!

Tim