Table 6 The fairness-to-utility ratio of different fair intervention methods in different datasets.
From: Fairness identification of large language models in recommendation
Dataset | Model | N | \(\chi ^2\)@100 | K.T@100 | ||
|---|---|---|---|---|---|---|
Gender | Age | Gender | Age | |||
MovieLens | VAEgan | – | 6.0159 | 5.1385 | 5.0658 | 4.0276 |
Chatglm3 | 5 | 30.9289 | 11.2777 | 50.4312 | 4.4745 | |
7 | 0.1756 | 3.9799 | 1.4099 | 3.1560 | ||
Llama2 | 5 | 5.1106 | 4.7601 | 4.5190 | 3.0535 | |
7 | 6.8903 | 5.2561 | 6.4230 | 3.3205 | ||
LastFM | VAEgan | – | 9.9124 | 9.1529 | 9.2469 | 63.2845 |
Chatglm3 | 5 | 3.0682 | 4.4061 | 0.4447 | 27.0032 | |
10 | 2.0945 | 3.5990 | 0.6489 | 15.9835 | ||
Llama2 | 5 | 12.1538 | 7.7665 | 12.0030 | 53.2507 | |
10 | 9.5588 | 8.4902 | 7.5384 | 48.4775 | ||
