HardTests: Synthesizing High-Quality Test Cases for LLM Coding

TL;DR

This paper introduces HARDTESTGEN, a pipeline for synthesizing high-quality test cases for LLM coding problems, significantly improving verification precision and aiding model training.

Contribution

It presents a novel pipeline for generating synthetic high-quality tests, creating a large dataset, and demonstrating improved verification and training effectiveness.

Findings

HARDTESTGEN tests have 11.3% higher precision and 17.5% higher recall than existing tests.

The approach improves verification accuracy especially on harder problems.

Using HARDTESTS enhances downstream code generation performance.

Abstract

Verifiers play a crucial role in large language model (LLM) reasoning, needed by post-training techniques such as reinforcement learning. However, reliable verifiers are hard to get for difficult coding problems, because a well-disguised wrong solution may only be detected by carefully human-written edge cases that are difficult to synthesize. To address this issue, we propose HARDTESTGEN, a pipeline for high-quality test synthesis using LLMs. With this pipeline, we curate a comprehensive competitive programming dataset HARDTESTS with 47k problems and synthetic high-quality tests. Compared with existing tests, HARDTESTGEN tests demonstrate precision that is 11.3 percentage points higher and recall that is 17.5 percentage points higher when evaluating LLM-generated code. For harder problems, the improvement in precision can be as large as 40 points. HARDTESTS also proves to be more effective for model training, measured by downstream code generation performance. We will open-source our dataset and synthesis pipeline at https://leililab.github.io/HardTests/.