测试大模型在知识任务中响应简单指令的能力，发现性能普遍大幅下降。
KCIF: Knowledge-Conditioned Instruction Following
- 用修改文本、数值、列表等指令改造知识问答题，测试模型响应能力。
- 所有模型性能显著下降，小模型降幅超80%，大模型也达40%-50%。
- 揭示知识与指令遵循应联合评估，适合模型评测与对齐研究者参考。
LLM评估基准传统上将知识/推理能力与指令遵循能力分开测试。本文研究知识与指令遵循的交互关系，发现大模型在执行简单答案修改指令时表现不佳，且容易被无关指令干扰。我们基于现有基于多选题的知识评测基准，施加一系列简单指令，包括文本改写（如改变大小写）、数值操作（如增加数值、改变格式）、列表操作（如排序答案项）以及干扰指令（如改变数值答案的大小写）。在不同参数规模（1B-405B）和模型族的大模型上进行评估，结果令人意外：所有模型在这些简单任务组合上的表现均出现显著下降。大型及前沿模型性能下降40%-50%，而中小型模型降幅更为严重（有时超过80%）。结果表明，传统上将知识/推理与指令遵循分离存在局限性，二者应联合研究。我们发布了基准数据集、评估框架代码及结果，供后续研究使用。
原文摘要 · Abstract (English)
LLM evaluation benchmarks have traditionally separated the testing of knowledge/reasoning capabilities from instruction following. In this work, we study the interaction between knowledge and instruction following, and observe that LLMs struggle to follow simple answer modifying instructions, and are also distracted by instructions that should have no bearing on the original knowledge task answer. We leverage existing multiple-choice answer based knowledge benchmarks and apply a set of simple instructions which include manipulating text (eg.: change case), numeric quantities (eg.: increase value, change formatting), operate on lists (eg.: sort answer candidates) and distractor instructions (eg.: change case of numeric answers). We evaluate models at varying parameter sizes (1B-405B) from different model families and find that, surprisingly, all models report a significant drop in performance on such simple task compositions. While large-sized and frontier models report performance drops of 40-50%, in small and medium sized models the drop is severe (sometimes exceeding 80%). Our results highlight a limitation in the traditional separation of knowledge/reasoning and instruction following, and suggest that joint-study of these capabilities are important. We release our benchmark dataset, evaluation framework code, and results for future work.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。