LIFT-Bench TRIAL

For teams training models to write reports, plans or documents, section-scoped constraints are closer to production than whole-answer style checks. Judge dependence and first-party results make this a reproducible trial, not a general capability claim.

Next stepWithin one day, pin a public split and the Gemma judge, then measure scoped-constraint pass rate, output length and judge consistency on one internal long-form template.

Judgment reviewed 2026-10-27 · Basis: AI21 opens LIFT-Bench for scoped long-form instruction following