IFBench

Measures: The model must satisfy verifiable output constraints it has not seen before (58 new ones such as keyword counts, formatting and sentence manipulation) on WildChat-style prompts. Code-based checkers score it, with no LLM judge. The headline is prompt-level loose accuracy.

Maintainer: Allen Institute for AI (Ai2)
Introduced: 2025
Status: Active
Saturation: Near-ceiling
Typical provenance: Mixed
In scoring set: Yes

Paper

Leaderboard

Show independent results only

Every result we hold, whoever ran it. Only independent and third-party rows count towards a score.

No results to show.