Upload README.md with huggingface_hub

README.md

@@ -7,6 +7,7 @@ tags:
 language:
 - multilingual
 base_model: jinaai/jina-reranker-v3
+base_model_relation: quantized
 inference: false
 license: cc-by-nc-4.0
 library_name: mlx
@@ -14,15 +15,7 @@ library_name: mlx
 
 # jina-reranker-v3-mlx
 
-MLX implementation of [jina-reranker-v3](https://huggingface.co/jinaai/jina-reranker-v3), a 0.6B parameter multilingual document reranker optimized for Apple Silicon.
-
-## Features
-
-- 🚀 Native Apple Silicon acceleration via MLX
-- 🎯 100% accuracy match with original PyTorch implementation
-- 📦 Minimal dependencies (no transformers needed)
-- 🌍 Multilingual support (same as original model)
-- ⚡ Efficient inference on M-series chips
+MLX implementation of [jina-reranker-v3](https://huggingface.co/jinaai/jina-reranker-v3), a 0.6B parameter multilingual document reranker optimized for Apple Silicon. Features native Apple Silicon acceleration via MLX with 100% compatibility to the original PyTorch implementation. No transformers library required.
 
 ## Installation
 
@@ -101,21 +94,6 @@ reranker = MLXReranker(
 )
 ```
 
-## Model Files
-
-This directory should contain:
-- `model.safetensors` - MLX-converted Qwen3 model weights
-- `projector.safetensors` - MLP projector weights
-- `tokenizer.json` - Tokenizer configuration
-- `config.json` - Model configuration
-- Other tokenizer files (vocab.json, merges.txt, etc.)
-
-## Performance
-
-Tested on Apple M-series chips with 100% ranking accuracy compared to the original PyTorch implementation:
-- Mean score difference: < 0.001
-- Perfect ranking matches: 100%
-- Inference speed: ~3-4s for 6 documents (Apple M1/M2)
 
 ## Citation