One good way of validating the RUSLE result is through the help of average seasonal suspended load of the watershed and seasonal discharge of the river. From which the total suspended load can be calculated which on adding the bedload (as per the nature of river through literature) can help to compare total sediment load with the output from RUSLE equation.

Article Assessment of soil erosion by RUSLE model using remote sensi...

Perhaps you would be interested in work by Dissmeyer and Foster 1980. I personally have not tried to validate its equations.

https://www.google.com/search?q=dissmeyer+and+foster&ie=UTF-8&oe=UTF-8&hl=en-us&client=safari#ip=1

Marion and Clingenpeel included a section in this paper on our use RUSLE to estimate effects.

https://www.srs.fs.usda.gov/pubs/gtr/gtr_srs161/gtr_srs161_308.pdf

i used a paper by Dissmeyer and Stump, 1980 for help in applying the RUSLE in our work in SE USA. Dissmeyer also developed methods and a paper that used filter fabric fences as a monitoring tool, and I customized the hillslope approach and applied to small gullies (Sediment from a small ephemeral gully, ~Hansen and Law, 2004 in researchgate). The problem with RUSLE is it addresses soil loss type erosion, and not necessarily soil delivery at catchment or watershed scales.

The WEPP model required too much information, when I was exploring methods to estimate erosion associated to forest management proposals in the the 1980s and 1990s, so I stuck with RUSLE and sediment delivery ratio estimates by Roehl (1962 I think). I uploaded an example of an EIS (2013) that used the RUSLE and drainage basin response units (small perennial catchments) to estimate effects from activity intensity, and determining when mitigation was appropriate.

Richard Burns, retired Hydrologist was part of the data collection team for the publication Dissmeyer and Stump 1980. It has been years, but he tried to explain to me that slope length is not an easy term. I believe he was trying to describe was the slope distance to where rill or gully erosion begins. Once rill and gully erosion occurs, you are outside of the bounds where the USLE was to be used as concentrated overland flow was occurring.

The WEPP model was developed to not only address erosion, but also include and address concentrated flows in rills, gullies, channels. With the USLE, the complex part goes from the equation developed from many small, well designed agricultural data collections and analysis under a wide variety of conditions and soils, to expanding this data to landscapes, watersheds, etc., and recognizing where the RUSLE applies, and where site changes and concentrated flow changes to soil loss storage or to accelerated soil movement.

In my years of RUSLE use, I have always tried to present numbers from RUSLE calculations as an estimate, a comparative tool from which to estimate generalized effects and alternative courses of activity or action, or need for special attention or mitigation. Important to be clear on what you want to validate.

I would like to add, I have used the filter fabric fences described by Dissmeyer, and for erosion of slopes, especially in forest conditions, we seldom caught enough sediment to feel like we were getting an accurate measure using standard surveying methods. Expanding to small catchment scale, there has been a much greater chance of accumulating and measuring sediment, though some of it is associated with concentrated flow. In an agricultural field setting, where soils are regularly exposed and subjected to soil loss type erosion, the filter fabric fences can capture conditions of substantial sediment movement so it can be accumulated and measured through time.

Dissmeyer, George E.; Stump, Richard F. 1978. Predicted erosion rates for forest management activities and conditions sampled in the Southeast. [Atlanta]: U.S. Department of Agriculture Forest Service, State and Private Forestry, Southeastern Region. 26 p

This is is the Dissmeyer and Stump paper referred to. If you want it and cannot find, I will see if I still have a scanned digital copy in my records.