Release Notes
Welcome to the release notes section. Here you’ll find an executive summary of recent updates across the entire LastMile platform. Click on a release to view the full details.
Nov 7, 2024
Introducing LastMile AutoEval
We are excited to introduce AutoEval, the first AI evaluator model fine-tuning service.
Features and Capabilities
- Out-of-the-box evaluation metrics: Key metrics for common AI application types, such as RAG and multi-agent systems, are available for immediate use.
- alBERTa evaluator models: 400M parameter NLI entailment models that are optimized for evaluation tasks like faithfulness, relevance and other key metrics.
- LLM Judge labeling service: Label your LLM application traces using LLM Judge evaluation criteria, and refine the labels with human-in-the-loop active labeling.
- Fine-tuning service: Design your own evaluation metric using the fine-tuning service. This is the first-of-its-kind service which you can use to develop as many evaluators as you need to measure AI application performance across a fully customized criteria.