MineValiCoder: Reliable Code Generation with Test Case Quality Mining and Bipartite Graph-Based Mutual Validation
MineValiCoder claims stronger code generation by filtering bad generated tests before using them to refine and rank code.
The paper presents a closed-loop TDD framework for cases where only natural-language requirements are available. It uses test-case self-validation, parallel code refinement, and a bipartite code-test scoring model to reduce misleading feedback from faulty or mixed-quality tests. Across four LLMs and several benchmarks, the authors report Pass@1 results of 96.34% on HumanEval, 87.40% on MBPP, 64.00% on APPS, and 51.33% on LiveCodeBench. ArXiv · AI/CL/LG's note
The paper presents a closed-loop TDD framework for cases where only natural-language requirements are available. It uses test-case self-validation, parallel code refinement, and a bipartite code-test scoring model to reduce misleading feedback from faulty or mixed-quality tests. Across four LLMs and several benchmarks, the authors report Pass@1 results of 96.34% on HumanEval, 87.40% on MBPP, 64.00% on APPS, and 51.33% on LiveCodeBench. ArXiv · AI/CL/LG's note
score 5