Core Concepts

TAR 2.0

Definition

A continuous TAR workflow (usually CAL) where training and review happen together until a stopping point is reached.

Why it matters

It finds responsive documents faster but shifts the defensibility question to the stopping decision.

In practice

In a TAR 2.0 workflow, review starts right away. The model ranks the collection, reviewers code the top-ranked documents, and every decision goes back into the model, which re-ranks what is left. Responsive documents tend to surface early, and the rate of new finds falls as review goes on.

Because training never stops, rolling collections are easier to handle: new documents are simply ranked alongside the rest.

The hard question moves to the end. The team needs a defensible reason to stop, usually a falling responsive rate in recent batches followed by a random validation sample of what was not reviewed.

Compare with

  • TAR 1.0: TAR 1.0 trains first, then classifies; TAR 2.0 learns from every review decision.
  • Continuous Active Learning (CAL): CAL is the specific learning protocol most TAR 2.0 tools use. The terms are often used interchangeably.
← Back to glossary