Is add_eq2 correct?

[blackcherry88]

Should it be self.add_expr([(a, n), (b, -m)])?

def add_eq2( self, a: str, b: str, m: float, n: float, dep: pr.Dependency ) -> None: # a/b = m/n if not self.add_expr([(a, m), (b, -n)]): return [] self.register2(a, b, m, n, dep)

[unhandyandy]

Where is this code?

[bughunter244]

Looks like it is here: https://github.com/google-deepmind/alphageometry/blob/main/ar.py#L425

I agree, it seems the dev makes a mistake there

[tpgh24]

Indeed this is a bug. I created a fork AG4Masses that fixed many issues in AlphaGeometry, including this one. Also added some additional features, documentation and idea on how to improve AG, check it out if you are interested.

I tested using a simple problem:

tt_rconst_bug
a b c = triangle12; d = midpoint a c ? cong a b a d

There is another bug related to constant ratio problems discussed in issue #111. To run the test problem correctly that bug also needs to be fixed.