Upload README.md with huggingface_hub

README.md

@@ -4,34 +4,11 @@ tags:
 - grpo-tool-saturation
 - sft
 - llamafactory-ready
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
-dataset_info:
-  features:
-  - name: k
-    dtype: int64
-  - name: r
-    dtype: int64
-  - name: tool
-    dtype: string
-  - name: prompt
-    dtype: string
-  - name: response
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 1333244
-    num_examples: 8000
-  download_size: 283618
-  dataset_size: 1333244
 ---
 
 # grpo-tool-sat-sft-corpus-v1
 
-SFT-only view of grpo-tool-sat-dataset-v1. 8000 demonstration rows with prompt/response columns ready for LlamaFactory. Overlap map-fraction 0.594; token-level shuffled across r-classes.
+SFT-only view of grpo-tool-sat-dataset-v1. v1.1 — prompt now ends with newline to match RL eval_prefix concatenation.
 
 ## Dataset Info
 
@@ -42,19 +19,20 @@ SFT-only view of grpo-tool-sat-dataset-v1. 8000 demonstration rows with prompt/r
 
 | Column | Type | Description |
 |--------|------|-------------|
-| k | Value('int64') | Key integer in [0, 9999] |
-| r | Value('int64') | k mod 3 (0 = M_only, 1 = T_only, 2 = Overlap) |
-| tool | Value('string') | Tool used in this demo (map or table) |
-| prompt | Value('string') | User-side prompt: Key: <k> |
-| response | Value('string') | Target assistant completion: prose + <tool_call> + <observation> + <answer> |
+| k | Value('int64') | Key integer |
+| r | Value('int64') | k mod 3 |
+| tool | Value('string') | map or table |
+| prompt | Value('string') | User prompt: "Key: <k>
+" (trailing newline) |
+| response | Value('string') | Target completion: prose + <tool_call> + <observation> + <answer> |
 
 ## Generation Parameters
 
 ```json
 {
-  "script_name": "src/data_gen.py (post-filter)",
+  "script_name": "src/data_gen.py + filter",
   "model": "n/a",
-  "description": "SFT-only view of grpo-tool-sat-dataset-v1. 8000 demonstration rows with prompt/response columns ready for LlamaFactory. Overlap map-fraction 0.594; token-level shuffled across r-classes.",
+  "description": "SFT-only view of grpo-tool-sat-dataset-v1. v1.1 \u2014 prompt now ends with newline to match RL eval_prefix concatenation.",
   "hyperparameters": {
     "seed": 1,
     "overlap_skew_map": 0.6,