TDFlow: Agentic Workflows for Test Driven Software Engineering

2025-10-27

AI Summary

To address the challenges of manual test interpretation and low repair success rates in warehouse-scale program repair, this paper proposes TDFlow: a multi-agent workflow for test-driven software engineering. TDFlow decomposes the repair task into four specialized subtasks—patch generation, debugging, revision, and optional test generation—executed iteratively by coordinated, large language model–driven agents constrained by a curated toolchain. This decoupling significantly reduces contextual overhead and enhances precision across each subtask. Evaluated on SWE-Bench Lite and Verified benchmarks, TDFlow achieves pass rates of 88.8% (+27.8% absolute improvement) and 94.3%, respectively. Manual validation confirms an exceptionally low test deception rate, demonstrating TDFlow’s effectiveness and robustness under realistic, human-curated test scenarios.

Technology Category

• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Multiagent Systems: Teamwork
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

We introduce TDFlow, a novel test-driven agentic workflow that frames repository-scale software engineering as a test-resolution task, specifically designed to solve human-written tests. Given a set of tests, TDFlow repeatedly proposes, revises, and debugs repository-scale patches using precisely engineered sub-agents and tightly constrained tools. The workflow decomposes software engineering program repair into four components governed by respective sub-agents. This simple, forced decoupling of patch proposing, debugging, patch revision, and optional test generation (1) reduces long-context burden on any individual sub-agent, (2) focuses each sub-agent on specific, pre-defined sub-tasks, and (3) allows for specialized performance improvement on specific sub-tasks. When provided human-written tests, TDFlow attains 88.8% pass rate on SWE-Bench Lite (an absolute improvement of 27.8% over the next best system) and 94.3% on SWE-Bench Verified. Manual inspection of the 800 TDFlow runs within SWE-Bench Lite and Verified uncover only 7 instances of test hacking, which were subsequently counted as failures. Furthermore, we show that the primary obstacle to human-level software engineering performance lies within writing successful reproduction tests. We envision a human-LLM interactive system powered by TDFlow where human developers write tests solved by LLM systems. Together, these results indicate that modern LLMs, when embedded in a narrowly engineered, test-driven workflow, already achieve human-level test resolution -- with the final frontier for fully autonomous repository repair being the accurate generation of valid reproduction tests.

Problem

• Solving repository-scale software engineering via test-driven workflows
• Automating patch proposal and debugging using specialized sub-agents
• Addressing human-level performance gaps in reproduction test generation

Innovation

• Agentic workflow frames software engineering as test-resolution
• Decomposes repair into four specialized sub-agent components
• Uses tightly constrained tools to propose revise debug patches