# Selector Score Breakdown

Selector v0.2 emits per-model score details from `scripts/run_selector_demo.py`.

Current demo input:

```text
团队每次都说要长期主义，但一到季度 KPI 就回到短期动作，这到底是什么问题？
```

Current output summary:

| Model | Score | Decision | Reasons | Penalties |
| --- | ---: | --- | --- | --- |
| `qpi` | 0.44 | selected | pipeline_position matched; complexity signal: 长期; selection_priority=9 | none |
| `intellectual_archaeology` | 0.02 | rejected | complexity signal: 长期; selection_priority=7 | no IA heavy-depth gate |

Rules represented:

- Negative trigger first.
- No-call threshold.
- QPI-before-IA gate.
- Rejected model reasons.
- No LLM selector.