Add dataset card

README.md

@@ -1,28 +1,56 @@
 ---
-dataset_info:
-  features:
-  - name: messages
-    list:
-    - name: role
-      dtype: string
-    - name: content
-      dtype: string
-  - name: source
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 56959628
-    num_examples: 23800
-  - name: test
-    num_bytes: 510700
-    num_examples: 200
-  download_size: 57225198
-  dataset_size: 57470328
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
-  - split: test
-    path: data/test-*
+license: mit
+task_categories:
+- text-generation
+tags:
+- instruction-tuning
+- sft
+- synthetic
 ---
+
+# Day 3 instruction blend
+
+A normalized multi-source instruction dataset built for the Day 3
+'instruction tuning at scale' exercise of a post-training curriculum.
+Every row is in the OpenAI-messages format with a `source` column for
+per-source ablations. Built by `day03_instruct/build_blend.py`;
+converters live in `common/format_convert.py`.
+
+- Blend size: 24000 rows (200 held out as `test`)
+- Sampling seed: 42
+- Length cap: 8000 total content characters per conversation
+
+## Sources
+
+| source | dataset | original format | weight | rows kept |
+|---|---|---|---|---|
+| alpaca | [yahma/alpaca-cleaned](https://huggingface.co/datasets/yahma/alpaca-cleaned) | alpaca | 1.0 | 8000 |
+| slimorca | [Open-Orca/SlimOrca](https://huggingface.co/datasets/Open-Orca/SlimOrca) | sharegpt | 1.0 | 8000 |
+| ultrachat | [HuggingFaceH4/ultrachat_200k](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k) | messages | 1.0 | 8000 |
+
+## Filtering
+
+Rows were rejected for structural problems (wrong role order, missing or
+non-final assistant turn, empty content, unknown speakers) or for exceeding
+the length cap. Per-source rejection counts:
+
+```json
+{
+  "alpaca": {},
+  "slimorca": {
+    "length: over 8000 chars": 96
+  },
+  "ultrachat": {
+    "length: over 8000 chars": 1902
+  }
+}
+```
+
+## Licensing and provenance
+
+- **alpaca** (yahma/alpaca-cleaned): CC BY 4.0; responses originally distilled from OpenAI models
+- **slimorca** (Open-Orca/SlimOrca): MIT; GPT-4-generated responses (note OpenAI-terms provenance caveat)
+- **ultrachat** (HuggingFaceH4/ultrachat_200k): MIT
+
+Portions of this data were generated by large language models; check the
+upstream dataset cards and the relevant providers' terms before commercial use.