Unverified50% confidenceFactExact time
In code generation evaluation, pass@1 measures the probability that the model passes the test with a single generation, while pass@10 allows the model to make 10 attempts and takes the best result
1
Sources
50%
Confidence
Long-term
Relevance
9/8/2026
First Seen
Sources
Related Entities
Related Claims
Unverifiedpass@1指标最初从代码生成评测领域（如HumanEval基准）借鉴而来，用于衡量模型一次生成正确代码的概率74% similarUnverifiedpass@k指标衡量模型生成k个候选代码片段时至少有一个能通过全部测试用例的概率74% similarUnverified每个模型配置以HIGH和XH（Extra High）两种推理强度进行测试，每种配置独立运行10次65% similarUnverifiedThe most reliable way to determine whether a new model has gotten worse is to conduct an A/B comparison between the old and new versions on the same tasks with the same prompts.62% similarUnverified团队在项目期间经历四五次模型迭代，每次都发现模型能力越强从验证循环中获得的收益越大，对验证循环的需求是更持久的底层规律61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/875814API
curl https://kongchang.com/api/v1/knowledge/claims/875814MCP
get_claim(id=875814)