Upload README.md with huggingface_hub

README.md

@@ -1,31 +1,39 @@
 ---
-dataset_info:
-  features:
-  - name: id
-    dtype: string
-  - name: section_text
-    dtype: string
-  - name: label
-    dtype: string
-  - name: is_synthetic
-    dtype: bool
-  - name: tier
-    dtype: int64
-  - name: agreed
-    dtype: bool
-  - name: tier_1_label
-    dtype: string
-  - name: tier_2_label
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 205568
-    num_examples: 200
-  download_size: 100088
-  dataset_size: 205568
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
+tags:
+- classify-and-augment
+- llm-annotated
 ---
+
+# davanstrien/otb-cascade-phaseB-2tier
+
+LLM-annotated dataset produced by [classify-and-augment](https://github.com/davanstrien/classify-and-augment).
+
+## Configuration
+
+- **Model**: Cascade: `LiquidAI/LFM2.5-1.2B-Instruct` → `Qwen/Qwen3-4B-Instruct-2507`
+- **Labels**: `jim_crow`, `no_jim_crow`
+- **Input rows**: 200
+- **Output rows**: 200
+
+## Label distribution
+
+| Label | Real | Synthetic | Total |
+|---|---:|---:|---:|
+| `jim_crow` | 14 | 0 | 14 |
+| `no_jim_crow` | 186 | 0 | 186 |
+
+## Cascade audit
+
+| Tier | Model | Processed | Entropy | Parse fail % | `jim_crow` count | `no_jim_crow` count |
+|---|---|---|---|---|---|---|
+| 1 | `LiquidAI/LFM2.5-1.2B-Instruct` | 200 | 0.515 | 0.0 | 23 | 177 |
+| 2 | `Qwen/Qwen3-4B-Instruct-2507` | 200 | 0.366 | 0.0 | 14 | 186 |
+
+**Pairwise tier agreement:**
+
+- Tier 1 vs Tier 2: 82.5% (165/200 examples)
+
+**Escalation:** 35 examples (17.5%) had cross-tier disagreement; tier-3 arbitration not used (no tier-3 configured; tier-2 wins).
+
+See [CASCADE_DESIGN.md](https://github.com/davanstrien/classify-and-augment/blob/main/CASCADE_DESIGN.md) for the disagreement-driven cascade strategy.
+