Defects4J

Defects-4J is a benchmark dataset that contains real-world Java programs with known defects, used to evaluate the effectiveness of code generation and repair techniques.

Papers using Defects4J (3)

BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models2026Reinforcement Learning-Guided Chain-of-Draft for Token-Efficient Code Generation2025RePaCA: Leveraging Reasoning Large Language Models for Static Automated Patch Correctness Assessment2025