HumanEval Pro and MBPP Pro: Evaluating Large Language Models on Self-invoking Code Generation

TL;DR

This paper introduces self-invoking code generation, a new evaluation task for LLMs that tests their reasoning by solving a base problem and then using its solution to solve a more complex one, along with new benchmarks.

Contribution

It proposes a general method to create challenging benchmarks for self-invoking code generation and analyzes LLM performance, revealing limitations and failure modes.

Findings

LLMs perform well on traditional benchmarks but less so on self-invoking tasks.

Instruction-tuned models show only marginal improvements over base models.

Performance drops significantly on self-invoking benchmarks compared to traditional ones.

Abstract

We introduce self-invoking code generation, a new task designed to evaluate the progressive reasoning and problem-solving capabilities of LLMs. In this task, models are presented with a base problem and a related, more complex problem. They must solve the base problem and then utilize its solution to address the more complex one. This work features three key contributions. First, we propose a general recipe for generating more challenging versions of existing benchmarks, resulting in three new benchmarks: HumanEval Pro, MBPP Pro, and BigCodeBench-Lite Pro, specifically designed to assess LLMs on self-invoking code generation. Second, from the analysis of experimental results over twenty LLMs on our benchmarks, we have two important observations: (i) Most LLMs excel in traditional code generation benchmarks like HumanEval and MBPP, but their performance declines on self-invoking tasks. For example, o1-mini achieves 96.2% pass@1 on HumanEval but only 76.2% on HumanEval Pro. (ii) On self-invoking code generation task, the instruction-tuned models demonstrate only marginal improvements compared to the base models. Third, we disclose the types of failure modes that exist in our evaluation results. All these results underscore the need for further advancements in self-invoking code generation tasks and provide a new direction for future research on enhancing LLMs' code reasoning capabilities.