Fable
There were two historians: one only checked whether it matched the original text, the other checked whether he had added his own words
The Hanlin Academy had two archive rooms: one collected external court bulletins and local memorials, and one kept the academy's own draft notes. The two historians sent out for field verification each had their own rules. Historian Zhao, after writing each passage, first opened the court bulletins and memorials to check—border grain prices, garrison organization, river water levels—whatever matched was called "stands up." Historian Qian did not look at the court bulletins; he only looked at the cited original text marked in his own draft notes—the draft said the border army had six battalions, but the original text only wrote four, so the extra two battalions were "added his own words."
First consolidated draft review: each of the two drafts has one passage flagged as anomalous. Historian Zhao's draft says "the river level rose to five chi two cun." Checking against the court gazette, the gazette indeed says five chi two cun—the fact is anchored. But the original annotated draft said "the water level was abnormally high," and Historian Zhao changed "abnormally high" to "five chi two cun"—the number matches the gazette, but not the meaning of the original. Historian Qian's draft left the number unchanged, but added a passage: "the water level was abnormally high, suspected upstream dam breach." The four words "suspected dam breach" appear in no source whatsoever.
In the second verification, the two historians swapped subjects. Historian Qian took Historian Zhao's draft to check against the court bulletin — all the numbers were accurate, full marks for grounding. Historian Zhao took Historian Qian's draft to check against the original text — faithfulness passed, with nothing added that was not in the original. The results of the two verification reports were completely different: the same draft was a fine piece by the grounding standard, but only half-rough by the faithfulness standard.
The scholar presiding over verification at the Hanlin Academy discovered the most dangerous case. Historian Zhao's draft had numbers that matched the court gazette, but the original memorial cited in the gazette had its numbers altered by the chief editor—the chain of anchoring was broken; the upstream was false data, and no matter how accurate the downstream numbers were, they were still wrong. Historian Qian's draft, though it added no extra content, had certain numbers that were actually vague descriptions in the original text calibrated by Historian Zhao using the court gazette—Historian Qian's fidelity standard could not see this source.
The scholar made a decision: from then on, verification would have two gates—the first gate, there must be no extra fabrication between the draft and the given source text; the second gate, the specific numbers and names cited in the draft must be traceable to external sources. Drafts that failed the first gate were returned directly; only after passing the first gate would they go through the second.
At the year-end overall review, the scholar wrote a summary: grounding checks whether "what you said is consistent with the external record," while faithfulness checks whether "you added things on your own to the materials given to you earlier." The two do not measure the same thing—a pair of matching drafts may be unfaithful to the source, and a draft faithful to the source may not match the facts. Quality evaluation needs two gates, walked in different orders.