I just tested this out. Here's my tips.

1. Be sure to list all the tags needed to activate the character as intended. "b3zthany" just generates a human women using whatever data wasn't associated to "furry female, bright purple hair" and all the other tags you put in there.

2. The tags female, perfect anatomy, well proportional body, anthro, female furry, anthro fox, three-tone fur, furry fox, white palms, long fluffy tail, and fluffy fur may not actually exist in the dataset. Illustrious was trained on Danbooru tags, which set the standard in image tagging, and you may search their site for more valid options.

Perfect anatomy and well proportional body are quality tags to be used with loras that some people just imitated and placebo-ed themselves into thinking it works. Tags like wide hips, curvy, and narrow waist exist that do meaningfully impact generations, so I suggest looking around there for what you want, assuming you would like those qualities to not be inherent to the activation tag b3zthany.

As for the other tags, here is my version of the prompt you trained on using existing tags. "b3zthany, solo, 1girl, furry, furry female, fox, orange fur, white fur, purple hair, light blue highlights, purple eyes, long hair, fluffy tail, very long tail, fluffy". Illustrious is generally flexible enough to the point where it would understand what you roughly meant for some of those tags, but it's good to stick with what the checkpoint knows to avoid confusing the lora and accidentally training them as secondary activation tags.

3. I saw you only really did this in the demo prompts and not the training, but the tags "score_9,score_8_up..." etc were only really a thing in Pony Diffusion due to a training error and are not present in Illustrious. Other tags like "25 years old" do not exist in the dataset either, as you can check on Danbooru, and do not do anything on their own.

4. I saw you trained every image on the same set of tags. While I can not see the images you trained on, it's generally a better idea to train on a diverse array of images with appropriately different tags to help the model know how a character would look under different conditions. While Illustrious is, again, flexible, there are some conditions where you can trip it up.

For example, what if I wanted to seriously poof her fur up? Normally I would add "fluffy", but you trained all images on the word, so the character is already "fluffy" to Illustrious regardless of what I may think. To fix this, you may forgo fluffy as a tag altogether, or only add it to images where she is more fluffy than usual.

You may trust an auto tagger to make these distinctions and check it's work later. Alternatively, you can try your own perception check to tag every image for their categories of backgrounds, poses, clothing, expressions, ear directionality (ears up or ears down), tail movement (like wagging), or other qualities. I have recommendations for programs you can use for automatic or manual taggers if you want to used them.

5. Be sure to check out guides for anything else you may be curious about, like training settings. I made an old guide myself on this that still holds up for the most recent models, but you can defiantly see what other people have to think.

Hope this helps