Value Occurrence Check

Step 2: Occurrence Scan

For every unique level energy in the table:

1. Use explicit Ei rows as the canonical level registry; do not cluster or average nearby energies
2. Collect all rows where the level appears as Ei or as derived final energy Ef = Ei − Eγ
3. For row-wise transition tables, match derived Ef to the nearest canonical Ei within ≤1.0 keV
4. If two candidates are equally near, flag the row as ambiguous rather than choosing one silently
5. Compare explicit energy strings character-for-character across repeated Ei occurrences
6. Compare uncertainty strings only if the table explicitly provides level uncertainties
7. Compare J-π character-for-character

Step 3: Report Inconsistencies

For each mismatch:

LEVEL: 1175.3 keV
  Row 7:  Ef=1175.3(1), Jπf=5/2-     ✓
  Row 26: Ei=1175.3(1), Jπi=5/2      ✗ parity mismatch (- vs none)
  Row 49: Ei=1175.3(2), Jπi=5/2-     ✗ uncertainty mismatch (1 vs 2)

Step 4: Validate Completeness

• Ground state (0 keV) appears only as Ef
• A missing explicit 0-keV row is not itself an inconsistency if the table encodes the ground state only through Ef≈0
• Every level energy has been checked
• Every data row has been checked against the canonical level registry
• 100% of occurrences verified before claiming completion

Data Fidelity

Preserve every decimal place exactly — do not round, omit, alter, or add digits. 10.0 remains 10.0, not 10 or 10.00. For energy matching: allow ≤1.0 keV tolerance when (Ei − Eγ) ≠ canonical Ei due to measurement precision.

Spot-Check Protocol

Use reproducible random sampling: seed (e.g., 12520260409), sample size ≥15% of rows, verify energy-matching residuals and J-π consistency on each sample. Record for each sampled row: source line, Ei, Eγ, derived Ef, matched canonical level, residual, and PASS/FAIL.

Gotchas

• Blank cells and non-data rows cause misalignment — exclude and count meticulously
• Markdown pipe tables often repeat headers between sections — do not count those as data rows
• AI models frequently fail at table corners — apply extra scrutiny
• Energy residuals from rounding are expected; focus on J-π consistency
• Near-degenerate levels can be distinct physical states — never merge them just because they are close in energy
• Normalize Unicode variants only for parsing if needed; keep reported source values unchanged
• Terminal (non-depopulated) levels do not indicate errors