add xyz model

[fangxiaoquan]

Checklist

• My model has a model sheet, report or similar
• My model has a reference implementation in mteb/models/ this can be as an API. Instruction on how to add a model can be found here
  • No, but there is an existing PR ___
• The results submitted is obtained using the reference implementation
• My model is available, either as a publicly accessible API or publicly on e.g., Huggingface
• I solemnly swear that for all results submitted I have not on the evaluation dataset including training splits. If I have I have disclosed it clearly.

[KennethEnevoldsen]

Results for fangxq/XYZ-embedding

task_name	fangxq/XYZ-embedding	google/gemini-embedding-001	intfloat/multilingual-e5-large
CmedqaRetrieval	0.48	nan	0.29
CovidRetrieval	0.91	0.79	0.76
DuRetrieval	0.91	nan	0.85
EcomRetrieval	0.70	nan	0.55
MMarcoRetrieval	0.83	nan	0.79
MedicalRetrieval	0.68	nan	0.51
T2Retrieval	0.86	nan	0.76
VideoRetrieval	0.81	nan	0.58
Average	0.77	0.79	0.64

task_name	Qwen/Qwen3-Embedding-8B	fangxq/XYZ-embedding
CmedqaRetrieval	0.53	0.48
CovidRetrieval	0.88	0.91
DuRetrieval	0.91	0.91
EcomRetrieval	0.73	0.70
MMarcoRetrieval	0.87	0.83
MedicalRetrieval	0.65	0.68
T2Retrieval	0.89	0.86
VideoRetrieval	0.80	0.81
Average	0.78	0.77

CmedqaRetrieval looks especially suspicious, this seems to be because the model was trained on this dataset (medqav2) - @Samoed I assume this PR will fix that? - if not, we should update the ModelMeta.

Otherwise, the model generally only performs surprisingly in cases where it was fit on the data, so once the annotation is resolved, we can merge.

[Samoed]

Yes, PR should fix this.