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)?
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