Dashboard

Overview of the ArbitrAI dashboard — your central hub for monitoring evaluation activity.

Dashboard

Overview

The Dashboard is the first thing you see after logging in. It gives you a quick overview of your workspace’s activity under the heading “Your work at a glance.”

Latest Benchmark Run

The left panel shows your most recent evaluation run with:

  • Document preview — a thumbnail of the evaluated document
  • Run source — e.g., “Key Discovery” for key abstraction runs
  • Created date
  • Summary cards highlighting:
    • Winner Model — model name and performance
    • Cheapest Model — model name and cost per document
    • Fastest Model — model name and latency

Two quick actions let you jump directly into results:

  • Inspect Model Outputs — opens the detailed per-document comparison view
  • View Results — opens the aggregate model comparison table

Usage and Activity

The right panel shows:

Evaluation Tokens

A bar chart displaying your token usage — how many evaluation tokens you’ve used out of your cap, and when the balance resets upon hover. Each token covers 1 document x 1 model x 1 run.

Recent Activity

A chronological timeline of recent events in your workspace:

  • Benchmark runs completed
  • Ground truth labels updated
  • Datasets created
  • Documents uploaded

Each event shows what happened and when.

Data Overview

At the right panel, summary counters give you a snapshot of your workspace:

  • Datasets — total count
  • Documents — total count
  • Evaluations — total benchmark runs
  • Prompts — saved prompt templates
  • Models — configured AI models
  • Label Coverage — percentage of documents that have ground truth labels
  • Model Providers — breakdown by provider with latest model added