Assessing and Improving the Representativeness of Code Generation Benchmarks Using Knowledge Units (KUs) of Programming Languages -- An Empirical Study

TL;DR

This study analyzes the representativeness of code generation benchmarks using Knowledge Units (KUs) and proposes a framework to augment benchmarks, leading to more realistic evaluations of large language models' coding abilities.

Contribution

It introduces the first empirical analysis of KU coverage in benchmarks, identifies their limitations, and presents a KU-based task synthesis framework to improve benchmark representativeness.

Findings

Benchmarks cover only half of the identified KUs.

Real-world projects exercise all KUs with balanced distribution.

Augmented benchmarks significantly reduce LLM performance overestimations.

Abstract

Large Language Models (LLMs) such as GPT-4, Claude and LLaMA have shown impressive performance in code generation, typically evaluated using benchmarks (e.g., HumanEval). However, effective code generation requires models to understand and apply a wide range of language concepts. If the concepts exercised in benchmarks are not representative of those used in real-world projects, evaluations may yield incomplete. Despite this concern, the representativeness of code concepts in benchmarks has not been systematically examined. To address this gap, we present the first empirical study that analyzes the representativeness of code generation benchmarks through the lens of Knowledge Units (KUs) - cohesive sets of programming language capabilities provided by language constructs and APIs. We analyze KU coverage in two widely used Python benchmarks, HumanEval and MBPP, and compare them with 30 real-world Python projects. Our results show that each benchmark covers only half of the identified 20 KUs, whereas projects exercise all KUs with relatively balanced distributions. In contrast, benchmark tasks exhibit highly skewed KU distributions. To mitigate this misalignment, we propose a prompt-based LLM framework that synthesizes KU-based tasks to rebalance benchmark KU distributions and better align them with real-world usage. Using this framework, we generate 440 new tasks and augment existing benchmarks. The augmented benchmarks substantially improve KU coverage and achieve over a 60% improvement in distributional alignment. Evaluations of state-of-the-art LLMs on these augmented benchmarks reveal consistent and statistically significant performance drops (12.54-44.82%), indicating that existing benchmarks overestimate LLM performance due to their limited KU coverage. Our findings provide actionable guidance for building more realistic evaluations of LLM code-generation capabilities.