[TriLUG] Spamassassin and auto-learn

[Joseph Tate]

I've been trying to get SpamAssassin to lower the auto-learn threshold, 
but haven't been able to do it successfully.

In my .spamassassin/user_prefs I have

bayes_auto_learn_threshold_nonspam  0.1
bayes_auto_learn_threshold_spam     10.0
bayes_auto_learn            1

But I frequenly have messages above 10, and sometimes as high as 18 that 
don't get fed through.

My criteria to validate whether a message has been run through 
auto-learn so far has been autolearn=spam in the X-Spam-Status header. 
Am I checking the wrong thing?

Does anyone have this feature working?