Full execution traces of GPT-OSS 120B running SWE-Agent on SWE-Bench (Full).
Each JSON file in traces/ corresponds to one SWE-Bench problem instance. The filename is the instance ID (e.g., django__django-12345.json).
{
"instance_id": "django__django-12345",
"model": "GPT-OSS-120B",
"agent": "SWE-agent",
"total_steps": 15,
"total_run_duration_seconds": 120.5,
"exit_status":… See the full description on the dataset page:
https://huggingface.co/datasets/zanderjiang/gpt-oss-120b-SWE-Agent.
