Unverified50% confidenceTradeoffExact time
Research settings typically allow multiple samplings and problem-type-specific prompt design, so researchers distinguish between the pass@k metric and the single-attempt pass rate.
1
Sources
50%
Confidence
Long-term
Relevance
9/23/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDifferences in evaluation setups across laboratories (such as pass@k, chain-of-thought prompting, and code execution tools) cause the same model to show different scores in different reports, complicating cross-laboratory comparisons.74% similarUnverifiedEvidently 内置了 KS 检验、卡方检验、PSI 等多种统计检验方法来量化漂移程度，并根据特征数据类型自动选择合适的检验方法70% similarUnverifiedResearchers are calling for the establishment of staged evaluation standards that map different levels of evidence to defensible claims of corresponding strength, avoiding over-generalization of limited experimental results70% similarUnverifiedEvaluating a replacement tool should involve small-scale controlled tests, running both the original and the replacement on identical research tasks to compare token consumption and output quality.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/943020API
curl https://kongchang.com/api/v1/knowledge/claims/943020MCP
get_claim(id=943020)