metadata
tags:
- classify-and-augment
- llm-annotated
davanstrien/otb-cascade-phaseB-2tier
LLM-annotated dataset produced by 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 for the disagreement-driven cascade strategy.