2024-03-28
What’s New
Pre-Joined Evals in Arize
Arize now supports LLM assisted evals that have been generated by the Arize Phoenix evals package. Use evals to determine the performance of your LLM application across dimensions such as Hallucination, Toxicity, QA Correctness and more. Evals can also be run on a job and sent to Arize on a regular cadence. See our docs here to get started with Evals in Arize, with more releases coming to Evals soon.
Enhancements
deleteData Endpoint
This update allows users to self-serve data deletion through GraphQL. Learn more →Area Under the Curve (AUC) as a Custom Metric
We now support AUC in custom metrics. Learn more →Python SDK v7.12.0
-
Users can now send evals and spans together via the
log_spansmethod of the Arize PandasClient - On-prem users can pass a path to certificate files or disable the TLS verification.
📚 New Content
The latest paper readings, ebooks, self-guided learning modules, and technical posts:- Tutorial: Everything You Need to Set Up a SQL Router Query Engine for Text-To-SQL
- LLM Task Evaluations vs Model Evals
- Anthropic Claude 3: Performance and Review
- Cerebral Valley on “How Arize Is Expanding the Field of AI Observability”
- Paper Read: Reinforcement Learning In An Era of LLMs