- 智能提取处理 (IXP) 发布说明
- Communications Mining 发行说明
- 非结构化文档和复杂文档发布说明
- 2025 年 9 月
- 2025 年 8 月

智能提取处理 (IXP) 发布说明
Introducing Gemini
Gemini is now available as an extraction model option. For more details, check Model configuration.
Model-based attribution
The Model Configuration panel now includes the Model-based attribution option. This provides improved attribution performance, with slightly increased latency compared to the existing default Rules-based attribution option. For more details, check Model configuration.
错误修复
- In the document annotation view, the document table filters and order you set did not persist when annotating and navigating through documents.
- Modified field values sometimes resulted in a mismatch indicator, although the value was the same.
- The Prompt override setting in the Model Configuration panel remained enabled, even after disabling it and saving the changes.
- The Monetary Quantity data type did not successfully parse numerical values with trailing minus signs, for example,
123-
. Instead, it returned these values as NaN, which means not a number. The data type now ignores the trailing minus sign and captures the number in front of it.
Introducing draft annotations
You can now save any changes you make to values and references as draft annotations before submitting them to the model. As a result, the document statuses appear as Partially annotated. For more details, check Draft annotations.
错误修复
- When manually editing a predicted text value, the text cursor jumped to the end of the string after the first edit.
- The project and validation scores showed zero while validation was partially complete. This issue occurred due to the lower number of annotated documents in the impacted projects.
Project score display updated
When a project score is unavailable, the label now shows Unavailable in grey text instead of Poor in red text.
Unsaved changes protection
A warning is now shown if the browser is closed while there are unsaved changes.
非结构化文档和复杂文档的推出
UiPath™ 智能提取处理 (IXP) 功能(非结构化文档和复杂文档的生成式提取)已在 GxP 美国地区正式发布。有关此功能的更多详细信息,请访问《非结构化文档和复杂文档指南》。
改进
字段名称现在最多可包含 128 个字符。
错误修复
展开提取内容时,视图会跳转到列表的末尾,需要您向上滚动到顶部。