0
9.0.4
Australia Patch 6, Australia Patch 5, Australia Patch 4, Australia Patch 3, Australia Patch 2, Australia Patch 1, Australia, Zurich Patch 12, Zurich Patch 11, Zurich Patch 10, Zurich Patch 9, Zurich Patch 8, Zurich Patch 7, Zurich Patch 4, Zurich Patch 2, Zurich Patch 1, Zurich, Yokohama Patch 11, Yokohama Patch 8, Yokohama Patch 6, Yokohama Patch 3, Yokohama
Generative AI
AI Data Kit is a one-stop solution for all your AI data needs. It empowers users to add, discover, curate, generate and selectively compile datasets, while seamlessly integrating with AI Skill Kit for skill evaluation. Users can attach ground truth to selected records through labeling workflows, include them in data collections, and publish these collections for use within Skill Kit during evaluation. In addition to data curation, Data Kit also enables users to generate synthetic data, making it a comprehensive platform for AI data preparation and management.
- Synthetic data generation to evaluate AI Agents and Skills
- Data scanning and cleansing to remove PII from datasets
- Data curation and management
- Labeling workflows
- Ground Truth generation and review flow for Agentic Evaluation
New
- Ground Truth creation: a new end-to-end flow to create, label, and review ground truth data for agentic evaluation, including a dedicated labelling experience with task and project views, and guided entry from the dataset page
- Datasets can now be added directly as an input source to labelling projects, with a configurable data sink step
- Ground truth data is now available for Skill Kit to consume directly in evaluation runs
- Data discovery through automatically extracted dataset metadata
- Record-level filtering within a dataset
- Japanese language support for Voice Evaluation data generation, with testing across all supported languages
Changed
- Improved synthetic data generation quality by incorporating agent instructions from Skill Kit
- Dataset creation now captures source details
Fixed
- Multiple defect fixes across the ground truth and labelling experience, the dataset page, and synthetic data generation
- Security fixes
Not applicable for this application version.
None