The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →If a harness logged “duplicates not canonicalized” because a duplicate query returned five rows, that result may be measuring page size—not duplicate handling. In the scale run described by Pranab Sarkar, every duplicate query requested top_k=5, and the harness counted returned rows with len(items) without checking whether their contents repeated a fact. Worse, the run’s planted duplicates were never written. The result did not show that canonicalization failed; it left that question unanswered.
Why did the harness report “duplicates not canonicalized”?
It treated the number of returned items as evidence of duplicate content. Sarkar’s account says the harness collected results for Q-dup-* queries, averaged each query’s returned value, and marked canonicalization false when that average reached two. But returned was just the number of items in the response: len(items). The code did not compare result text or identities to determine whether any facts were duplicates. Read the author’s explanation of the metric.
As an Amazon Associate I earn from qualifying purchases.
Because each query requested top_k=5, a full page could produce an average of 5.0 whether its rows contained duplicates or not. That number describes how many results came back, not what those results said.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →The example makes the mismatch clear: a query aimed at a preferred “nord” color scheme returned five loosely related editor preferences. None was the target fact, and none duplicated another result. Five rows were returned, but that was not evidence of five matching facts—or of failed duplicate canonicalization.
#1 Best Overall
Why did the write shortfall make the verdict even weaker?
The probe attempted to write 1,000 facts but, according to Sarkar’s article, reported about 256 successful writes before stalling. The planted duplicate records were IDs 901–1000, beyond the portion the run had stored. A query cannot establish how the system handles test records that were never present in its store. See the reported run counts and corpus layout.
That leaves two distinct defects:
- The duplicate metric was invalid: it counted returned rows instead of checking whether results repeated facts.
- The intended duplicate cases were absent: the run stopped far short of writing the records that were meant to test canonicalization.
The appropriate conclusion is not that canonicalization works, nor that it failed. This run did not test that claim, so the verdict is unknown.
What failed in the write phase?
Sarkar describes the write stall as a real software issue, separate from the bad duplicate verdict. He attributes it to Python constructors that created an engine but did not start its background worker pool; without the compactor, he says, the in-memory delta tier filled at 256. The article says an engine change dated June 27, 2026 fixed the issue and that the plugin’s minimum engine version at publication was 0.12.1. These are the author’s account of the cause and fix; they do not turn the duplicate result into a valid failure finding. Read the author’s explanation of the write stall.
What does the separate 16/20 score establish?
The article also reports precision-at-five of 16/20 on a separate retrieval check over facts that had been written successfully. That score is not evidence about duplicate canonicalization. Sarkar notes that the check used word-overlap scoring on a templated corpus and had not been audited hit by hit, so it should be read as a limited, author-reported result—not as a general benchmark or independent validation. See the qualification on the retrieval score.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to tell whether a benchmark verdict is supported
For each reported result, check that the evidence actually measures the stated behavior and that the relevant test inputs were present.
- Check the metric against the claim. To test duplicates, inspect result identities or normalized content for repeated facts. A returned-row count only measures response size.
- Check what was successfully written. Distinguish planned records from stored records, and confirm the specific cases behind a metric landed before interpreting their results.
- Check the raw response. A verdict should be traceable to the underlying output that supports it, rather than only to a summary number.
- Report missing coverage as unknown. If the run never wrote the target inputs, it cannot establish how the system handled them.
Repeatability does not repair a bad measurement: a harness can reliably return five rows on every top-five query while never checking whether those rows duplicate one another. Sarkar puts the reporting standard plainly: “A verdict should cite the raw response that backs it, and a verdict computed from a run that never wrote its test inputs should say unknown.” Read the author’s conclusion.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
