Unverified50% confidenceDecision RuleExact time
For growth scenarios, evaluating an agent should also incorporate "experiment win rate" — how many of the merged changes achieve statistically significant positive results in A/B tests.
1
Sources
50%
Confidence
Long-term
Relevance
11/3/2026
First Seen
Sources
Related Entities
Related Claims
Unverified对同一测试用例执行多次调用统计成功比例得到通过率，比二元的通过/失败判断更能反映模型真实表现77% similarUnverifiedIn a natural experiment, 7 of 11 skill assessment scores changed between two runs, with a maximum fluctuation of 1 point.72% similarUnverifiedMeta AI research shows that having a model iteratively optimize its own prompts can improve performance by 15-30% on certain standard benchmarks, but carries approximately a 20% probability of falling into a degenerative loop.71% similarUnverifiedAccording to the study's description, this processing noticeably improves the model's accuracy and consistency of judgment on samples involving binary payloads—the blind spot is partially filled.71% similarUnverifiedMacro-average accuracy first computes the accuracy of each class separately and then takes the arithmetic mean, making it more sensitive to the performance of minority classes; overall accuracy is the proportion of all samples predicted correctly and is easily affected by the class sample size distribution.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1129735API
curl https://kongchang.com/api/v1/knowledge/claims/1129735MCP
get_claim(id=1129735)