Are probabilities given by GIDEON tested for accuracy?

Yes, several studies have assessed GIDEON's ranking of potential diseases.

Links to published studies follow.

2008 https://pubmed.ncbi.nlm.nih.gov/18310530/
2006 https://pubmed.ncbi.nlm.nih.gov/16323112/
2005 https://pubmed.ncbi.nlm.nih.gov/17292033/
2001 https://pubmed.ncbi.nlm.nih.gov/11485667/
1999 https://pubmed.ncbi.nlm.nih.gov/10381973/
1998 https://pubmed.ncbi.nlm.nih.gov/9524864/

Note that studies largely assessed the ranking of diseases: 'Did the disease listed at the top of the differential diagnosis list reflect the true diagnosis?'

GIDEON does not make such a claim. In fact, the disease listed at the top of the list only carries a higher statistical probability than other diseases on the list.