If you have held out a test set and got similar Sharpe and correlation on that, then I’d say those are excellent. In the end, all that matters is live performance and if you can get positive Sharpe and mean correlation consistently, I’d say that’s good too. Ideally different from the meta model and what others have already discovered.