yeah, agree that that's a nice relaxation of this problem. this work focuses on property inference:

→ View original post on X — @jxmnop