# Research Evidence Ledger · 研究证据账

> Use when claims, sources, novelty, and confidence start collapsing into one vague score. Keep
> evidence strength separate from paradigm distance.

## 1. Claim Register

| Claim | Source | Evidence grade | Method / basis | Boundary | Link |
|---|---|---|---|---|---|
|  |  | I / II / III / IV / V |  |  |  |

Grades:

- **I**: law, direct measurement, or audited primary record.
- **II**: peer-reviewed empirical result with clear method.
- **III**: preprint, theory, or strong but not yet settled result.
- **IV**: practitioner report, de-identified case, or expert account.
- **V**: exploratory inference, analogy, or speculation.

## 2. Believability × Distance

| Claim | Evidence strength | Paradigm distance | Cell | Action |
|---|---|---|---|---|
|  | weak / strong | near / far | doubt / integrate / suspend / human |  |

Do not kill `weak × far` claims automatically. Suspend them and ask what evidence would discriminate
reframing from noise.

## 3. Integration Queue

| Item | Why it matters | Connects to | Human synthesis needed | Due |
|---|---|---|---|---|
|  |  |  |  |  |

## 4. Falsifiers

| Claim | What would prove it wrong? | Where to watch | Review date |
|---|---|---|---|
|  |  |  |  |

## 5. Output Rule

No generated research artifact counts until it is either:

- integrated into the evidence base,
- suspended with a discriminating test,
- or rejected with a recorded reason.