揭示大型语言模型及其基准测试中的能力差距
Uncovering Competency Gaps in Large Language Models and Their Benchmarks
AI总结：
提出一种基于稀疏自编码器概念激活的新方法，自动发现模型在细粒度概念上的弱点（模型差距）和基准测试覆盖不平衡（基准差距），并通过内部表示评估和跨基准比较进行验证。
摘要：
大型语言模型的评估严重依赖标准化基准测试。这些基准测试提供了有用的聚合指标，但可能掩盖（i）模型薄弱的特定子领域（“模型差距”）和（ii）基准测试本身的不平衡覆盖（“基准差距”）。为了自动揭示这两类差距，我们提出了一种简单的新方法，利用稀疏自编码器的概念激活，在逐概念基础上识别细粒度差距。该方法还受益于将评估基于模型的内部表示，以及易于跨基准测试进行比较。我们将该方法应用于五个流行的开源模型和十几个基准测试，作为示例说明。作为对该方法的验证，我们发现我们的自动无监督方法能够恢复文献中先前记录的模型差距（例如与谄媚相关的差距），并识别出新的模型差距。我们还能够自动揭示基准差距：应属于给定基准测试范围的核心概念。我们的“能力差距”方法可以通过提供模型行为的概念级分解，并帮助基准测试开发者迭代基准测试设计，来补充现有基准测试。代码可在 https://competency-gaps.github.io 获取。
英文摘要：
The evaluation of large language models relies heavily on standardized benchmarks. These benchmarks provide useful aggregated metrics, but can obscure (i) particular sub-areas where the models are weak ("model gaps") and (ii) imbalanced coverage in the benchmarks themselves ("benchmark gaps"). To automatically uncover both types of gaps, we propose a simple new method using concept activations from sparse autoencoders, to identify fine-grained gaps on a per-concept basis. The method also benefits from grounding evaluation in the model's internal representations, as well as easy comparison across benchmarks. We applied the method to five popular open-source models and more than a dozen benchmarks, as illustrative examples. As validation of the approach, we found that our automatic, unsupervised method was able to recover model gaps that have been previously documented in the literature (e.g. relating to sycophancy), in addition to identifying novel model gaps. We were also able to automatically uncover benchmark gaps: core concepts that should fall within the scope of a given benchmark. Our "competency gaps" method can be used to complement existing benchmarks, by providing a concept-level decomposition of model behavior, and by helping benchmark developers iterate upon benchmark design. Code is available at https://competency-gaps.github.io.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
