Type: url
Original: https://arxiv.org/pdf/2405.18216
Status: generated
Summary
A survey that consolidates 327 studies on modern code review published between 2013 and 2025 into a taxonomy of improvement techniques and understanding studies, then uses a SWOT analysis to weigh the capabilities of LLM-based review against its context gap, hallucination and metric problems and its socio-technical risks. It proposes three paradigm shifts for future research: context-aware proactivity, value-driven evaluation, and human-centric symbiosis.