Rethinking Verification for LLM Code Generation: From Generation to Testing

TL;DR

This paper critically examines current LLM code evaluation methods, introduces a new collaborative testing approach called SAGA, and develops TCGBench to improve test coverage and reliability in code generation assessment.

Contribution

It proposes a human-LLM collaborative test-case generation method (SAGA) and a new benchmark (TCGBench) to enhance test coverage and evaluation accuracy for LLM code generation.

Findings

SAGA achieves a 90.62% detection rate on TCGBench.

Verifier accuracy of generated benchmarks increased by 10.78%.

Enhanced evaluation reliability for LLM code generation.

Abstract

Large language models (LLMs) have recently achieved notable success in code-generation benchmarks such as HumanEval and LiveCodeBench. However, a detailed examination reveals that these evaluation suites often comprise only a limited number of homogeneous test cases, resulting in subtle faults going undetected. This not only artificially inflates measured performance but also compromises accurate reward estimation in reinforcement learning frameworks utilizing verifiable rewards (RLVR). To address these critical shortcomings, we systematically investigate the test-case generation (TCG) task by proposing multi-dimensional metrics designed to rigorously quantify test-suite thoroughness. Furthermore, we introduce a human-LLM collaborative method (SAGA), leveraging human programming expertise with LLM reasoning capability, aimed at significantly enhancing both the coverage and the quality of generated test cases. In addition, we develop a TCGBench to facilitate the study of the TCG task. Experiments show that SAGA achieves a detection rate of 90.62% and a verifier accuracy of 32.58% on TCGBench. The Verifier Accuracy (Verifier Acc) of the code generation evaluation benchmark synthesized by SAGA is 10.78% higher than that of LiveCodeBench-v6. These results demonstrate the effectiveness of our proposed method. We hope this work contributes to building a scalable foundation for reliable LLM code evaluation, further advancing RLVR in code generation, and paving the way for automated adversarial test synthesis and adaptive benchmark integration.