Link released Qev-train synthetic dataset

README.md

@@ -3,18 +3,20 @@ license: apache-2.0
 base_model: Qwen/Qwen3.5-9B-Base
 base_model_relation: adapter
 language:
-  - en
-  - zh
+- en
+- zh
 tags:
-  - qev
-  - qwen3_5
-  - lora
-  - decision-model
-  - choice
-  - noul
-  - score
+- qev
+- qwen3_5
+- lora
+- decision-model
+- choice
+- noul
+- score
 metrics:
-  - accuracy
+- accuracy
+datasets:
+- AustinFu/Qev-train
 ---
 
 <p align="center">
@@ -90,14 +92,14 @@ python -m qev.predict \
 | Late mixing | Starts halfway through main training; late examples repeat three times |
 | Selected checkpoint | Step 2327 |
 
-The main and late partitions intentionally share 249 replay records. The complete research training corpus is not distributed with this release. [Training guide](https://github.com/QiqianFu/Qev/blob/main/docs/training.md) · [Data recipe](https://github.com/QiqianFu/Qev/blob/main/docs/data.md).
+The main and late partitions intentionally share 249 replay records. [Qev-train](https://huggingface.co/datasets/AustinFu/Qev-train) publishes 1,842 synthetic alignment, rule-compliance and world-knowledge examples with original hard labels and generation documentation. The complete mixed training corpus is not distributed. [Training guide](https://github.com/QiqianFu/Qev/blob/main/docs/training.md) · [Data recipe](https://github.com/QiqianFu/Qev/blob/main/docs/data.md).
 
 ## Evaluation
 
 **Qev-9B uses BF16 backbone computation; Kev-9B uses FP32.**
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/QiqianFu/Qev/main/assets/evaluation.svg" alt="Qev and Kev accuracy on seven benchmarks." width="100%">
+  <img src="https://raw.githubusercontent.com/QiqianFu/Qev/main/assets/evaluation.svg" alt="Qev-9B and Qev-2B accuracy compared with their Qwen3.5 base models on seven benchmarks." width="100%">
 </p>
 
 | Benchmark | Jev (reference) | Qwen3.5-9B-Base | Qev-9B | Kev-9B |