# mmau / b09b362f-110a-4829-b94a-8337cd1a906b - taskset: [mmau](https://harnessreport.com/tasks/mmau.md) - difficulty: medium - category: audio - language: - runnable from the site: no - agent timeout: 600s ## Results by harness _none yet_ ## Instruction ``` You are given a transcript of an audio clip. Transcript: [BEGIN TRANSCRIPT] The verbatim transcription of the spoken words in the audio clip is: 'Oh my God, guess what? Guess what? Guess what? What? Well, guess guess guess guess um you don't look like a man to me. No. Let me see. Oh, no way. You were posing, weren't you?' [END TRANSCRIPT] Question: [BEGIN QUESTION] What did First speaker tell Second speaker not to look at? [END QUESTION] Choices: [BEGIN CHOICES] Her left hand., Her right hand., Her face., Her left foot. [END CHOICES] Select the single best answer from the choices. Output requirements: - Write ONLY the exact answer text (must match one of the choices). - Create a file at /app/answer.txt. - The file must contain only the answer text, with no extra characters or whitespace. Example: If choices are ["Man", "Woman", "Child", "Robot"] and the correct answer is Woman, write: echo -n "Woman" > /app/answer.txt ``` --- Harness Report runs agent harnesses from their GitHub repos on Harbor tasks and records every model call. Every page is also `.md` and `.json`; index: https://harnessreport.com/llms.txt · MCP: https://harnessreport.com/mcp