For VXP models, it is best to increase the weight to around 1.0.
If you try using a model other than Animagine XL V3.1 and it doesn't look similar, try increasing the weight as well.
Learning model: AnimagineXL
This lora has only been tested with "Animagine XL V3.1" and "VXP_XL v2.2 (Hyper)". I don't know if it will work properly with models other than the above.