Unverified50% confidenceFactExact time
Median L1 error is the median of the errors taken over multiple test samples, which avoids individual difficult samples inflating the average.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe researchers found that when χ1 is close to 1, the test error decreases monotonically in all four experimental units.77% similarUnverifiedSetting an absolute failure count ceiling for small-sample slices is more robust than setting a percentage threshold and is insensitive to sample size76% similarUnverifiedA single call passing the test only gives an n=1 result, which is insufficient to prove model reliability; the pass rate and the lower bound of its confidence interval need to be computed over multiple trials.75% similarUnverifiedIn differential testing, a non-zero curl exit should fail the case outright, because two identical errors look like a match to the diff.75% 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.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114491API
curl https://kongchang.com/api/v1/knowledge/claims/1114491MCP
get_claim(id=1114491)