PRBench · Legal
Scale AI / Professional work · Answers complex legal-work questions to assess professional judgment and arguments.
What it measures and how
Fixes either Legal's full 500 tasks or Hard's 250, with common search tools, tiers, and judges. Uses the rubric aggregate without counting both sets or describing professional answers as complete file delivery.
How this evidence is used
Under review: could add legal judgment. Tasks, tools, and batches need checking. Source listing only; no automatic collection or scoring.
Evaluation limits & attribution
Model judges introduce preferences. Legal and finance tasks from one organization share an influence boundary, rather than being independent evidence. This does not directly measure Word, Excel, or slide production.
Data license: Official tasks and code are public; licenses do not automatically cover live results. Stable exports and current-result redisplay terms are unconfirmed.
Results are published by Scale AI. Original scores and the NexGateHub consensus index use different scales and cannot be added directly.