Request for agent's raw traces

#11
by kyzor - opened

Hi Mercor Team,
Thanks for the APEX-Agents benchmark! I’m currently researching agentic reliability and error correction during long-horizon tasks.
Is there a way to access the raw trajectory traces or logs for the evaluated models (e.g., GPT-5, Claude, Gemini)?

Mercor org

Hey, @kyzor yes, this is something we offer to researchers on request. Please email me at apex@mercor.com, and we can arrange access. Please let us know your affiliation and intended use.

kyzor changed discussion status to closed

Sign up or log in to comment