Multi-IF

Multi-IF is a benchmark for how precisely a model obeys the instructions it was given. The official description says: "Multi-IF - multi-turn instruction following benchmark evaluating complex multi-constraint instruction adherence". Below we describe in simple words what benchmarks of this type generally examine.

Multi-IF measures how precisely a model obeys the instructions it was given. In detail, it checks:

• meeting verifiable constraints (length, format, language, layout)
• following several constraints correctly at once
• keeping the constraints over long, multi-turn conversations
• not ignoring parts of the instruction while sounding plausible

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better.

Official methodology note: Multi-IF - multi-turn instruction following benchmark evaluating complex multi-constraint instruction adherence.

What is the Multi-IF benchmark?

Multi-IF is a benchmark for how precisely a model obeys the instructions it was given. The official description says: 'Multi-IF - multi-turn instruction following benchmark evaluating complex multi-constraint instruction adherence'. Below we describe in simple words what benchmarks of this type generally examine.

What does Multi-IF test?

Multi-IF mainly measures how precisely a model obeys the instructions it was given. Concretely, it checks: meeting verifiable constraints (length, format, language, layout); following several constraints correctly at once; keeping the constraints over long, multi-turn conversations.

How is the Multi-IF score calculated?

The score is accuracy: the percentage of tasks answered or solved correctly. 100% means every task was solved perfectly; higher is better. The declared maximum score of this benchmark is 100.0.

Where do I get Multi-IF data and scores?

The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/multi-if/scores) or pull it directly from the Open LLM Radar API with any HTTP client.