Evaluated data libraries processing problems

The paper deals with issues related to the preparation of working libraries for applied tasks based on the files of the evaluated data. Cases of identifying contradictions that arose as a result of errors in data and processing programs are given. On examples of many years of practice, their quantitative and qualitative assessments are given. It is shown that distortions carried by errors in the processing process can increase the error of the given data, thereby leveling the efforts of experimenters, theoreticians, and evaluators spent on improving the accuracy of the data. The following are indicated as objective prerequisites for a relatively high probability of occurrence of errors: a continuous variety of options for presenting data; regular revisions of the formats for each new assessment; the need to configure programs for new data; the need to expand the functionality of programs for new applications. It is shown that the sources of uncertainty of the final results can be: the duality of data representation, the ambiguity of processing methods, the instability of solutions due to the loss of signs, the non-optimally chosen grid of reference points, the non-consistency of versions of libraries and programs, the inevitable errors in the representations of the estimated data and algorithms for their transformations. Due to the rather high frequency of detection of errors in files and programs, a list of measures aimed at their prompt elimination and reduction of the scatter of the final results due to distortion of information in the processing process is given. To implement these measures, it is proposed to create an administered resource that allows you to quickly implement feedback between evaluators-compilers of files, mathematicians-programmers and specialists in constants responsible for the preparation and validation of applied libraries.