Table 1. Sample Statistical Results of Transformer-Based and GRU-Based Generators.
pretrained model |
optimized model (aver reward = 15) |
|||
---|---|---|---|---|
transformer | GRU | transformer | GRU | |
invalid | 1.694% ± 0.102% | 4.368% ± 0.361% | 1.170% ± 0.052% | 2.690% ± 0.167% |
redundant | 0.075% ± 0.022% | 0.038% ± 0.015% | 30.273% ± 0.525% | 33.828% ± 0.616% |
retained | 98.231% ± 0.115% | 95.594% ± 0.351% | 68.557% ± 0.552% | 63.482% ± 0.456% |