SWE-Compass: Towards Unified Evaluation of Agentic Coding Abilities for Large Language Models
This addresses the need for a unified and production-aligned evaluation framework for agentic coding abilities in LLMs, benefiting researchers and developers, but it is incremental as it builds on existing benchmarks.
The authors tackled the problem of limited evaluation of large language models for software engineering by introducing SWE-Compass, a comprehensive benchmark that spans 8 task types, 8 scenarios, and 10 languages with 2000 instances, and they benchmarked ten state-of-the-art models, revealing a clear hierarchy of difficulty.
Evaluating large language models (LLMs) for software engineering has been limited by narrow task coverage, language bias, and insufficient alignment with real-world developer workflows. Existing benchmarks often focus on algorithmic problems or Python-centric bug fixing, leaving critical dimensions of software engineering underexplored. To address these gaps, we introduce SWE-Compass1, a comprehensive benchmark that unifies heterogeneous code-related evaluations into a structured and production-aligned framework. SWE-Compass spans 8 task types, 8 programming scenarios, and 10 programming languages, with 2000 high-quality instances curated from authentic GitHub pull requests and refined through systematic filtering and validation. We benchmark ten state-of-the-art LLMs under two agentic frameworks, SWE-Agent and Claude Code, revealing a clear hierarchy of difficulty across task types, languages, and scenarios. Moreover, by aligning evaluation with real-world developer practices, SWE-Compass provides a rigorous and reproducible foundation for diagnosing and advancing agentic coding capabilities in large language models.