← Back to context

Comment by adityamishra241

19 hours ago

Interesting approach. The iterative refinement + verifiable constraints combination seems especially useful. Curious what kinds of tasks have worked best in practice.

The approach seems to work reasonably well on tasks with evals or benchmarks (public or hand-made). We don't have a very large set of use cases yet, but we got good results on seemingly far-apart domains like game level design, predictive modeling and some toy problems on bioinformatics.