Update README.md

README.md

@@ -2,10 +2,16 @@
 license: mit
 library_name: transformers
 base_model: Qwen/Qwen3-Reranker-0.6B
-pipeline_tag: text-generation
+pipeline_tag: token-classification
 tags:
 - code
 - context-pruning
+- agent
+datasets:
+- Raymone023/SWE-QA-Benchmark
+metrics:
+- f1
+- mse
 ---
 
 # SWE-Pruner: Self-Adaptive Context Pruning for Coding Agents
@@ -20,6 +26,11 @@ Inspired by how human programmers selectively skim code, SWE-Pruner enables agen
 
 Evaluations across benchmarks show that SWE-Pruner achieves 23-54% token reduction on agent tasks like SWE-Bench Verified and up to 14.84x compression on single-turn tasks like LongCodeQA with minimal performance impact.
 
+## Model Usage
+Given that we have made significant modifications to the model, its dual-head architecture and the complex compression head service code will be rather complex. 
+Therefore, we recommend that you use the version we have released on [GitHub](https://github.com/Ayanami1314/swe-pruner) instead of attempting to use the original model on your own.
+
+
 ## Citation
 If you find SWE-Pruner useful in your research, please cite:
 ```bibtex