Sources & reliability

Where these claims come from, and how much to trust them. Every claim was checked by an LLM judge panel before it entered the graph; this page shows the source mix and the judges' confidence.

15documents

33043claims

6%repaired

0.98avg groundedness

Claims per source

COFFEE pg2850019616

SUGAR pg736165334

GENERAL pg507591203

GENERAL pg530271050

pg573401011

SPICES pg60192965

okakura book of tea867

pg24588772

TEA pg46775506

history of coffee501

SUGAR pg52820407

history of tea273

The corpus leans on a few sources — read every chart with that in mind.

Judge scores — groundedness

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.71950

0.8279

0.932869

How confident the judges were (0–1); accepted claims cluster high.

Judge scores — qualifier_fidelity

0.027

0.10

0.21

0.348

0.440

0.5530

0.61397

0.7311

0.83608

0.929136

How confident the judges were (0–1); accepted claims cluster high.