Lens divergence detected
Quick scan (+0.00)
vs full evaluation (+0.17)
· Δ 0.17
The holistic Llama estimate and section-level Claude evaluation disagree by more than 0.15.
The full evaluation is more reliable for individual story analysis.
SETL History
structural-editorial tension over time · 3 days