Upload merged Qwen3-4B-Instruct-2507 model (auto-generated README)

README.md

@@ -45,7 +45,7 @@ Loss is applied to **all assistant turns** in the trajectory, enabling the model
 
 - DBBench dataset: `u-10bei/dbbench_sft_dataset_react_v4`
 - ALFWorld dataset: `u-10bei/sft_alfworld_trajectory_dataset_v5`
-- Mixing ratio (pre-merge target): **DB:ALF = 1:1**
+- Mixing ratio (pre-merge target): **DB:ALF = 3:1**
 
 ### DB Oversampling (category-aware)
 Enabled: **False**
@@ -66,8 +66,8 @@ DB category weights used during training-data preparation:
 - Method: LoRA (full precision base)
 - Max sequence length: 4096
 - Epochs: 1
-- Learning rate: 8e-05
-- LoRA: r=32, alpha=64, dropout=0.0
+- Learning rate: 3e-05
+- LoRA: r=32, alpha=64, dropout=0.05
 - Per-device train batch size: 2
 - Gradient accumulation: 4
 

model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0d2d71802a7dcdff7594ba31701343454c03052c239272cc7ab62203b4f789df
+oid sha256:fd754441f37fcaeabebf5f7d200d69d88198dafbf7ade0c78b6e1aafd3164871
 size 4967215360

model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0d029aeebf305468eb1cdba3d8a44d06157f9e45d8918d58090276a17ae40fb1
+oid sha256:49fb62246505ed17e61b85f1dc5de3f3e325b94da7aca2a1db3d2622a29fc9e1
 size 3077766632