Survey: LLM-based agents for software engineering
Broad 2024 survey of LLM agents across SE tasks—useful map of roles, tools, and feedback loops including review/testing.
Source: Large Language Model-Based Agents for Software Engineering: A Survey
This survey catalogs how LLM agents are applied across software engineering—planning, coding, testing, and review-like roles—and what scaffolding patterns recur. It is a bibliography with a spine: agents extend LLMs with tools, memory, and multi-step control.
Use it as a map, not as primary evidence. When a new review agent paper drops, this survey helps place it against prior multi-agent SE systems.
Why it matters for code review
Review is one node in an agentic SDLC graph. The survey is handy for spotting whether a “new” review architecture is actually a remix of planner/tester/reviewer roles the field has already tried.