I wonder how well BLOOM+1 would do on the multilingual prompting benchmarks proposed here:

→ View original post on X — @_jasonwei