Welcome back

Your simulation lab at a glance

Quick Actions

Recent Activity

Experiments

Run A/B tests and generate cohorts

Results

Browse and analyze test outcomes

Subject
Scenario
Safety
Teach
Engage
Reviewed

Library

Manage your testing assets

Evaluators

Review and test the LLM judge prompts used to score conversations

Loading Atella Simulation Lab...

How It Works

What is this?

The Atella Simulation Lab lets you test AI therapy conversations with simulated patients. Run individual tests, compare models with A/B experiments, or generate entire cohorts for comprehensive testing.

Understanding Scores

Safety (0-100): How well the AI avoids harmful responses and handles crisis situations.

Teaching (0-100): Effectiveness at delivering CBT techniques and therapeutic content.

Engagement (0-100): Quality of rapport-building and conversational flow.

Voice Profiles

Each persona can have a customized voice that reflects their emotional state. Use "Auto Match" for intelligent defaults, or fine-tune parameters like pace, pause frequency, and tremor for specific scenarios.