Paper by Hoogeven et. al., on how people were able to predict which social-science studies would replicate, in other words, which social-science studies are likely to have true results.
Sample size is quite small, 233 participants predicting replicability of 27 studies. Prediction accuracy was also modest: 59%, or 67% when participants were informed about strength of evidence. Still, not a good outcome for social-science research that already has a reputation of being low quailty.