SWE Atlas - Codebase QnA
coding
SWE Atlas - Codebase QnA evaluates a model's ability to answer questions about real codebases, measuring repository-level comprehension and the ability to reason about code structure, behavior, and intent across an entire project.
Methodology
Imported from llm-stats public benchmark metadata. Modality: text. Max score: 1. Categories: agents, code. Language: en. Verified by llm-stats: no.