[[["易于理解","easyToUnderstand","thumb-up"],["解决了我的问题","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["很难理解","hardToUnderstand","thumb-down"],["信息或示例代码不正确","incorrectInformationOrSampleCode","thumb-down"],["没有我需要的信息/示例","missingTheInformationSamplesINeed","thumb-down"],["翻译问题","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["最后更新时间 (UTC):2025-03-06。"],[[["This is a beta feature, subject to the \"Pre-GA Offerings Terms\" and available \"as is\" with potentially limited support, as detailed in the Service Specific Terms and launch stage descriptions."],["Feature attributions reflect the feature's impact on a specific prediction, not necessarily the model's overall behavior, requiring dataset aggregation for a broader understanding."],["Attributions are based solely on the model and its training data, only showing if the model is utilizing a feature, not if a relationship exists between a feature and a target."],["Improving attribution precision involves adjusting parameters like the number of integral steps or paths in methods like integrated gradients, XRAI, and sampled Shapley."],["The choice of attribution method depends on the image type; integrated gradients are suited for low-contrast or non-natural images, while XRAI works best with natural, higher-contrast images, but cannot handle low contrast, overly tall or wide, or very large images."]]],[]]