Communication regarding data sources and accuracy
#2
by redauzhang - opened
How are the current data sources and accuracy guaranteed? Have the injected samples generated by the large model been thoroughly tested? I had the large model directly generate injected black and white samples, and the accuracy was only 82%. Does this meet the model's expectations? For FN, there are a large number of Windows RFI samples with bypassed encoding.
redauzhang changed discussion status to closed