This study investigated the repeatability and correlations of social behavior in wild populations and showed their association with fitness. This manuscript is very well-written, and the dataset collected by the authors is comprehensive and potentially useful for future research. I have some major comments regarding the data filtering, statistical models, and result presentation.
1. Authors collected a large amount of data, but only a relatively small portion of it went into the formal statistical analysis. For example, the authors had almost 400 individuals but only around 200 of them were used in the analysis of social behavior. This is the result of data filtering. I am not sure whether the conclusion holds true if the authors used a larger dataset. I would perform sensitivity analysis but reduce the cut-off of at least 30 or 25 sightings to 10 to 15 sightings.
2. Authors used SNPs data to assign parents to estimate the fitness. However, the minor allele frequency threshold seems unusually high. Given the maximum possible allele frequency being 0.5, the authors set 0.43. In my experience, the usual MAF threshold is 0.05 or 0.1. This also reduced the number of SNPs to only 180 (from the original more than 2000).
3. The main concern lies in the structure of statistical models. The authors used Gaussian distributions for all the models (lines 234 to 337). However, because the data collected are proportional and count data, binomial and Poisson distributions will be an appropriate distribution. With that, authors should also consider over-dispersion. This will likely influence the conclusions drawn from the models, including repeatability and correlations. (I know authors reported the distributions of the residuals, but diagnostic plots should be the relationship between residual and fitness values and qq-norm plots, not the histogram of residuals).
4. The presentations of results could be improved. Many figures and tables are redundant. Specifically, figure 1 and table 2 are pretty much the same. Figure 2 and Table 3 are the same. Table 4, figure 3, figure 4, figure 5, and figure 6 are conveying the same information.

Minor comments
Line 23. It is not clear what “preference” is without context.
Line 45. I think the authors mean “between” individuals.
Line 127. How does DArTcap differ from RNA-seq?
Line 168. Add parentheses to the equation.
Line 206. Did the authors consider the quality score and linked equilibrium?
Line 365. How the social behavior changes across the year should also be reported.