Justin from Broadly Construed has a nice post (from a long time past) on evidence and I thought I'd post a thought I posted over there in the comment box.
Here's an argument against E = B. (It seems to work equally well against any view that identifies evidence with the contents of non-factive mental states):
(i) Inference to best explanation requires that evidence is constitutionally capable of figuring in explanation.
(ii) Evidence is constitutionally capable of figuring in explanation only if it is true.
(iii) Evidence consists of truths.
(iv) There are false beliefs.
(v) ~(E = B)
The case for (ii) is that explanations are factive. ('p because q' is true only if p and q are true). I don't think there's a second step in evaluation that determines whether something that is evidence can then serve as the evidence that figures in IBE. It seems (iii) follows from (ii).
It seems to me that this causes trouble for a host of views, particularly if it is combined with an acquisition/access principle to the effect that non-inferential knowledge is sufficient for the possession of evidence. Here's a gloss on what evidence is: it is whatever we can properly rely on in theoretical reasoning (of the sort that terminates in belief) without needing to rely on additional evidence or beliefs. I can't see how non-inferential knowledge fails to do that. I also can't see how non-inferentially justified beliefs could fail to do that. So, if we say that non-inferentially justified beliefs provide evidence and accept the argument above, we get the following:
(*) S's belief that p can be non-inferentially justified only if p.
That doesn't rule out the possibility of false, justified beliefs but it's a start.