SWE Atlas - Test Writing

coding

Multilingual: no

SWE Atlas - Test Writing evaluates a model's ability to author meaningful tests for real-world software projects, measuring how well agents can understand code and produce correct, useful test coverage.

Methodology

Imported from llm-stats public benchmark metadata. Modality: text. Max score: 1. Categories: agents, code. Language: en. Verified by llm-stats: no.