IFEval

Instruction-Following Evaluation for Large Language Models

541 个提示,每个包含 25 类可验证指令中的一种或多种,例如"至少写 400 字"、"以 JSON 回复"、"包含关键词 X 三次"或"不使用逗号"。合规性通过程序检查,给出严格与宽松匹配下的提示级和指令级准确率。低成本、可复现且不依赖 LLM 评判,是检验聊天模型是否真正遵守格式约束的标准。

95%
Qwen3.5-27B
厂商自报
2023-11-14 → 2026-02-24: 76.89% → 95%
发布
2023-11
维护者
Google (Zhou et al.)
状态
saturating
污染风险
high
指标
prompt-level strict accuracy (percent, ↑)
题量
541
领域
instruction-following
人类
无实测基线

备注. Vendors frequently report instruction-level or loose accuracy without saying so, which inflates numbers by 5-10 points; check the variant before comparing. Most frontier models exceed 90% prompt-level strict, and IFBench and IFEval-Hard have emerged as harder follow-ups.

完整账本

系统开发者分数日期来源条件
Qwen3.5-27BAlibaba95%厂商自报
Model card benchmark table; same table lists GPT-5-mini (2025-08-07) at 93.9 as run by Qwen. Variant not stated. Also the top row on benchlm.ai's IFEval page.
Qwen3-235B-A22B-Instruct-2507Alibaba88.7%厂商自报
Model card 'Alignment' table; the same table lists Kimi K2 at 89.8, Claude Opus 4 (non-thinking) 87.4 and GPT-4o-0327 83.9 as run by Qwen. Variant not stated.
Llama 3.1 405B InstructMeta88.6%厂商自报
Llama 3.1 model card benchmark table (70B: 87.5, 8B: 80.4); Meta does not state which IFEval variant. Date is the Llama 3.1 release.
GPT-4OpenAI76.89%论文split: prompt-level strict
Table 3 of the IFEval paper (responses collected November 2023); instruction-level strict 83.57, loose 79.30/85.37.