What it measures

Multi-level fine-grained constraint-following benchmark adding 1 to 5 incremental constraints per seed instruction.

Reading the measurement

Separate satisfying instructions from factual accuracy, human preference and safe behavior. Inspect strict all-constraint success as well as average criterion credit. Judge choice and adversarial prompt selection can materially change the result.

Published evaluation

Published mean over the five constraint difficulty levels. Hard and soft success rates are different measures and remain separate.

Where this fits in an evaluation suite

Format compliance, preference, truthfulness, and safety require separate measurements. Keep desirable writing style from masking factual errors.

Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.

A reproducible comparison

Published mean over the five constraint difficulty levels. Hard and soft success rates are different measures and remain separate.

Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.

PRIMARY EVIDENCE & FURTHER READING

  1. FollowBench: A Multi-level Fine-grained Constraints Following Benchmark for Large Language Models — published result table ↗

    Jiang, Yuxin, Wang, Yufei, Zeng, Xingshan, Zhong, Wanjun, Li, Liangyou, Mi, Fei et al. · 2023

    All 10 authors

    Jiang, Yuxin, Wang, Yufei, Zeng, Xingshan, Zhong, Wanjun, Li, Liangyou, Mi, Fei, Shang, Lifeng, Jiang, Xin, Liu, Qun, Wang, Wei

Reviewed 2026-10-26. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.