Unverified50% confidenceBenchmarkExact time
In the error-correction task, five models detect at least 96.7% of implanted errors, with a false-positive rate of no more than 4%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/4/2026
First Seen
Valid until: 2/2/2027
Sources
Related Entities
Related Claims
UnverifiedThe error rate of manual data entry is typically between 0.5% and 4%, and the error rate is higher for tasks involving complex judgment.79% similarUnverifiedA 5% Span sampling rate can capture only about 5% of the content of a failed run, because failures are rare and correlated, and the leaf Spans with real diagnostic value happen to be exactly the batch already flagged by the state filter.79% similarUnverifiedIn 20 public targeted test cases of "omitted inverse-square correction," the three best-performing models caught the error in every worksheet, while the three weaker models accepted 16 to 18 of the 20 worksheets with implanted errors.79% similarUnverifiedWhen monitoring the error rate of large model APIs, focusing only on the overall success rate is insufficient; attention should be paid to error rate percentiles and spike detection, because averages may mask the impact of tail error rates on some users.77% similarUnverifiedIn 11 of the 12 models, between 87.5% and 100% of capitulation cases stopped at the exact wrong number provided by the user, meaning capitulation typically involves switching sides entirely.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1134263API
curl https://kongchang.com/api/v1/knowledge/claims/1134263MCP
get_claim(id=1134263)