AI Code Review Study: Faster but Not Better
QueensUniversity · hf · 2026-07-20
This study analyzes 1.02 million code reviews (Pull Requests) across 207 GitHub projects, exploring the evolution across three phases: from "manual review" to "LLM-assisted review" and finally "agentic review".
Key Findings:
- Faster Reviews: Introducing AI agents (especially multi-Agent collaborations or Agent-initiated reviews) significantly accelerates review decision speeds.
- No Quality Improvement: This efficiency boost does not translate into higher code review quality.
- Paradigm Shift: Once LLMs and AI Agents are involved, human-AI collaboration becomes the strongest factor explaining review efficiency, lessening the importance of traditional variables like code activity types.
More from coding & agent
- Hugging Face wraps 6-month Training Agents series with open recordings and code — SergioPaniego · 2026-10-26
- Agent verified 'success' after double-voting: a lesson in harness vs model bugs — TejasKumar_ · 2026-10-26
- diginomica: agentic automation hype is overdone, and 'The Hackening' of enterprise security has begun — jonerp · 2026-10-26
- 'Vibe verification': ~60% of AI-written formal specs prove the wrong property — basedjensen · 2026-10-26
- Devin launches Security suite powered by Claude Mythos with $100M+ credits — DevinAI · 2026-10-26
- HackBench injects reward hacks into agent trajectories to stress-test monitors — DanielKhashabi · 2026-10-26