Test sets designed to measure how well an automated agent performs legal tasks, such as drafting, extraction, research or review.

Cited as though a shared standard exists, in a field where each vendor publishes its own test set, scores itself on it, and no two results in the category can be placed side by side.

• No Common Instrument: Every vendor's benchmark is different, which makes every vendor top of its own table.
• Graded by the Ungraded: The hardest part of legal work is judgment under uncertainty, which is the part no test set has managed to score.
• House-Cleaning Reality Check: Like every restaurant on the street publishing its own food ratings and each one coming first.

Love Island (Season 5): each islander explains their type on paper using criteria entirely their own, and no two of those standards can be measured against each other.

Test sets for measuring how well an agent performs legal tasks.

“We lead every major legal agent benchmark in the industry.”

“We scored 82 percent on a 200-task set we built ourselves, and no competitor has run the same set.”