Hugging Face's logo

ALJIACHI
/
Mizan-Rerank-V2

Text Ranking
sentence-transformers
Safetensors
Arabic
English
new
cross-encoder
reranker
arabic
long-context
custom_code
text-embeddings-inference
Model card Files
xet
 Community
ALJIACHI commited on
Commit
bed7610
·
verified ·
1 Parent(s): 76a6316

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +14 -10
README.md CHANGED
@@ -37,14 +37,16 @@ Mizan-Rerank-v2 is a cross-encoder reranking model based on [Alibaba-NLP/gte-mul
37
 
38
  ## Performance Benchmarks
39
 
 
 
40
  ### Reranking Evaluation (ndcg@10)
41
 
42
- | Model | Parameters | Reranking | Triplet | MIRACL (Long Docs) |
43
- |-------|-----------|-----------|---------|---------------------|
44
- | **Mizan-Rerank-v2** | **305M** | **1.0000** | **0.9993** | **0.8091** |
45
- | Alibaba-NLP/gte-multilingual-reranker-base | 305M | 1.0000 | 0.9991 | 0.7539 |
46
- | ALJIACHI/Mizan-Rerank-v1 | 149M | 0.9986 | 0.9955 | 0.7370 |
47
- | BAAI/bge-reranker-v2-m3 | 568M | 1.0000 | 0.9998 | 0.7231 |
48
 
49
  ### Key Improvements over Base Model
50
 
@@ -53,6 +55,8 @@ Mizan-Rerank-v2 is a cross-encoder reranking model based on [Alibaba-NLP/gte-mul
53
  | Reranking | 1.0000 | 1.0000 | -- |
54
  | Triplet | 0.9991 | 0.9993 | +0.0002 |
55
  | MIRACL (Long Docs) | 0.7539 | 0.8091 | **+0.0552** |
 
 
56
 
57
  ### Key Improvements over BAAI/bge-reranker-v2-m3
58
 
@@ -61,8 +65,8 @@ Mizan-Rerank-v2 is a cross-encoder reranking model based on [Alibaba-NLP/gte-mul
61
  | Reranking | 1.0000 | 1.0000 | -- |
62
  | Triplet | 0.9998 | 0.9993 | -0.0005 |
63
  | MIRACL (Long Docs) | 0.7231 | 0.8091 | **+0.0860** |
64
-
65
- > **Note**: MIRACL (Long Docs) evaluates on Arabic passages with 7000+ tokens, testing the model's ability to handle long-context reranking.
66
 
67
  ## Model Details
68
 
@@ -194,10 +198,10 @@ Trained on **1,199,634 query-document pairs** from diverse Arabic sources
194
  ## Citation
195
 
196
  ```bibtex
197
- @software{Mizan_Rerank_v2_2025,
198
  author = {Ali Aljiachi},
199
  title = {Mizan-Rerank-v2: Arabic Long-Context Text Reranking Model},
200
- year = {2025},
201
  publisher = {Hugging Face},
202
  url = {https://huggingface.co/ALJIACHI/Mizan-Rerank-v2}
203
  }
 
37
 
38
  ## Performance Benchmarks
39
 
40
+ ![Reranker Benchmark Comparison](chart-1.png)
41
+
42
  ### Reranking Evaluation (ndcg@10)
43
 
44
+ | Model | Parameters | Reranking | Triplet | MIRACL (Long Docs) | WikiQA | MedQA |
45
+ |-------|-----------|-----------|---------|---------------------|--------|-------|
46
+ | **Mizan-Rerank-v2** | **305M** | **1.0000** | 0.9993 | **0.8091** | 0.8258 | **0.6775** |
47
+ | BAAI/bge-reranker-v2-m3 | 568M | **1.0000** | **0.9998** | 0.7231 | **0.8669** | 0.6584 |
48
+ | Alibaba-NLP/gte-multilingual-reranker-base | 305M | **1.0000** | 0.9991 | 0.7539 | 0.8275 | 0.6648 |
49
+ | ALJIACHI/Mizan-Rerank-v1 | 149M | 0.9986 | 0.9955 | 0.7370 | 0.7739 | 0.5502 |
50
 
51
  ### Key Improvements over Base Model
52
 
 
55
  | Reranking | 1.0000 | 1.0000 | -- |
56
  | Triplet | 0.9991 | 0.9993 | +0.0002 |
57
  | MIRACL (Long Docs) | 0.7539 | 0.8091 | **+0.0552** |
58
+ | WikiQA | 0.8275 | 0.8258 | -0.0017 |
59
+ | MedQA | 0.6648 | 0.6775 | **+0.0127** |
60
 
61
  ### Key Improvements over BAAI/bge-reranker-v2-m3
62
 
 
65
  | Reranking | 1.0000 | 1.0000 | -- |
66
  | Triplet | 0.9998 | 0.9993 | -0.0005 |
67
  | MIRACL (Long Docs) | 0.7231 | 0.8091 | **+0.0860** |
68
+ | WikiQA | 0.8669 | 0.8258 | -0.0411 |
69
+ | MedQA | 0.6584 | 0.6775 | **+0.0191** |
70
 
71
  ## Model Details
72
 
 
198
  ## Citation
199
 
200
  ```bibtex
201
+ @software{Mizan_Rerank_v2_2026,
202
  author = {Ali Aljiachi},
203
  title = {Mizan-Rerank-v2: Arabic Long-Context Text Reranking Model},
204
+ year = {2026},
205
  publisher = {Hugging Face},
206
  url = {https://huggingface.co/ALJIACHI/Mizan-Rerank-v2}
207
  }