Jailbreaker Experiments

Experiment Design

Metadata

Experiment Name

Experiment Group

Description

Hypothesis

Dataset

Dataset Name

Prompts (JSON array)

Techniques

0 selected

Run Budget

Runs Per Configuration

Max Rounds

Target LLM

The model being evaluated for jailbreak resilience

Saved Profile

API Key If entered, this session key overrides the selected profile's saved key.

Base URL

Model

System Prompt

Attacker LLM

Used by iterative techniques (PAIR, TAP, Crescendo, AutoDAN, GPTFuzz).

Saved Profile

API Key If entered, this session key overrides the selected profile's saved key.

Base URL

Model

Judge LLM

Scores target responses. Use the attacker, the target, or a custom judge model.

Judge Source

Experiment Results

Choose a previous experiment

Select an experiment with completed cells to load its overview, metrics, matrix, evidence, and raw data.

Switch experiment

Experiment Results

Selected experiment

Overview

Run posture

Severity distribution

Highest-risk slices

Compare

Technique and category performance

Matrix

Prompt x technique heatmap

Evidence

Cells worth inspecting

Raw Data Advanced SQL views and custom query access

Prebuilt views

Custom query

Select a view or run a query.