SWE-Swiss: a training recipe for software engineering models
Researchers from Peking University, ByteDance Seed, and HKU present a training approach for software engineering tasks. The SWE-Swiss-32B model, with 32 billion parameters, scored 60.2% on SWE-bench Verified, according to the authors.
Researchers from Peking University, ByteDance Seed, and HKU present SWE-Swiss, a training recipe for software engineering tasks. According to the reported results, SWE-Swiss-32B, with 32 billion parameters, scored 60.2% on SWE-bench Verified.
The result may help engineers compare training approaches, but it does not by itself show that the recipe will suit different data or goals. When using AI to study or adapt the material, avoid submitting proprietary code, credentials, or personal data; use fictional or authorized examples and follow your organization’s policies.