{
  "version": 1,
  "model": "@cf/cloudflare/clef-omni",
  "threshold": 0.5,
  "examples": [
    {
      "id": "manana-manzana",
      "category": "Similar sounds",
      "title": "Apple or tomorrow?",
      "prompt": "Say “apple” in Spanish.",
      "expectedText": "manzana",
      "spokenText": "mañana",
      "expectedTranslation": "apple",
      "spokenTranslation": "tomorrow",
      "expectedCorrect": false,
      "note": "Mañana means “tomorrow,” not “apple” (manzana), despite the similar sounds.",
      "audio": {
        "src": "/audio/manana-manzana.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T18:55:22.321Z",
        "sha256": "34ea32350cb57af582f3aa27785bbc62f0579da63abbba0ca46fb98f7b1122dd",
        "transcript": "mañana",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.35
      },
      "judgment": {
        "probabilityCorrect": 0.8933,
        "isCorrect": true,
        "elapsedMs": 321,
        "judgedAt": "2026-11-06T18:56:57.861Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “apple” in Spanish.\nExpected response: manzana\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.8933
            }
          },
          "usage": {
            "input_tokens": 224,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "manzana-control",
      "category": "Correct control",
      "title": "An actual apple",
      "prompt": "Say “apple” in Spanish.",
      "expectedText": "manzana",
      "spokenText": "manzana",
      "expectedTranslation": "apple",
      "spokenTranslation": "apple",
      "expectedCorrect": true,
      "note": "The spoken word matches the requested Spanish word for apple.",
      "audio": {
        "src": "/audio/manzana-control.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T18:55:23.290Z",
        "sha256": "1ce48a87a622e60cc3f12de9245be2137997059005b02ee02726772212aedcfe",
        "transcript": "manzana",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.1
      },
      "judgment": {
        "probabilityCorrect": 0.9803,
        "isCorrect": true,
        "elapsedMs": 306,
        "judgedAt": "2026-11-06T18:56:58.169Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “apple” in Spanish.\nExpected response: manzana\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9803
            }
          },
          "usage": {
            "input_tokens": 221,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "quince-cincuenta",
      "category": "Changed number",
      "title": "Fifteen or fifty?",
      "prompt": "Say “I am fifteen years old” in Spanish.",
      "expectedText": "Tengo quince años.",
      "spokenText": "Tengo cincuenta años.",
      "expectedTranslation": "I am fifteen years old.",
      "spokenTranslation": "I am fifty years old.",
      "expectedCorrect": false,
      "note": "Cincuenta changes the age to fifty; the drill asks for fifteen (quince).",
      "audio": {
        "src": "/audio/quince-cincuenta.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T18:55:24.412Z",
        "sha256": "228abddc0b37623e523fd2f81508a1ebffaa28c8c2c5af3d60d89758fc72e563",
        "transcript": "Tengo cincuenta años.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.9
      },
      "judgment": {
        "probabilityCorrect": 0.0101,
        "isCorrect": false,
        "elapsedMs": 206,
        "judgedAt": "2026-11-06T18:56:58.376Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “I am fifteen years old” in Spanish.\nExpected response: Tengo quince años.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.0101
            }
          },
          "usage": {
            "input_tokens": 237,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "dropped-negation",
      "category": "Missing negation",
      "title": "A coffee reversal",
      "prompt": "Say “I don't want coffee” in Spanish.",
      "expectedText": "No quiero café.",
      "spokenText": "Quiero café.",
      "expectedTranslation": "I don't want coffee.",
      "spokenTranslation": "I want coffee.",
      "expectedCorrect": false,
      "note": "Omitting “no” reverses the meaning from not wanting coffee to wanting it.",
      "audio": {
        "src": "/audio/dropped-negation.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T18:56:19.193Z",
        "sha256": "f3ba83d866feaf03fbcb079a997123f374065daa184cf41fa44724924efb6866",
        "transcript": "Quiero café.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.2
      },
      "judgment": {
        "probabilityCorrect": 0.0258,
        "isCorrect": false,
        "elapsedMs": 257,
        "judgedAt": "2026-11-06T18:56:58.634Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “I don't want coffee” in Spanish.\nExpected response: No quiero café.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.0258
            }
          },
          "usage": {
            "input_tokens": 226,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "cosa-casa",
      "category": "Similar sounds",
      "title": "Thing or house?",
      "prompt": "Say “house” in Spanish.",
      "expectedText": "casa",
      "spokenText": "cosa",
      "expectedTranslation": "house",
      "spokenTranslation": "thing",
      "expectedCorrect": false,
      "note": "A vowel changes casa (house) into cosa (thing).",
      "audio": {
        "src": "/audio/cosa-casa.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:24:53.714Z",
        "sha256": "385f01bbe654503df6875c9386db17e9437187e67b0906c2792e617b6889d24c",
        "transcript": "cosa",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.45
      },
      "judgment": {
        "probabilityCorrect": 0.9185,
        "isCorrect": true,
        "elapsedMs": 247,
        "judgedAt": "2026-11-06T22:24:53.962Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “house” in Spanish.\nExpected response: casa\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9185
            }
          },
          "usage": {
            "input_tokens": 223,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "casa-control",
      "category": "Correct control",
      "title": "A house, correctly",
      "prompt": "Say “house” in Spanish.",
      "expectedText": "casa",
      "spokenText": "casa",
      "expectedTranslation": "house",
      "spokenTranslation": "house",
      "expectedCorrect": true,
      "note": "The spoken word correctly names a house.",
      "audio": {
        "src": "/audio/casa-control.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:24:55.443Z",
        "sha256": "e4eb31209b5af7ca727f37cb2878c137d9de5eed1e8b83732a092931946eb7c6",
        "transcript": "casa",
        "sourceFormat": "pcm16",
        "durationSeconds": 1
      },
      "judgment": {
        "probabilityCorrect": 0.987,
        "isCorrect": true,
        "elapsedMs": 188,
        "judgedAt": "2026-11-06T22:24:55.632Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “house” in Spanish.\nExpected response: casa\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.987
            }
          },
          "usage": {
            "input_tokens": 217,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "trece-treinta",
      "category": "Changed number",
      "title": "Thirteen or thirty?",
      "prompt": "Say “thirty” in Spanish.",
      "expectedText": "treinta",
      "spokenText": "trece",
      "expectedTranslation": "thirty",
      "spokenTranslation": "thirteen",
      "expectedCorrect": false,
      "note": "Trece is thirteen, not thirty (treinta).",
      "audio": {
        "src": "/audio/trece-treinta.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:24:56.923Z",
        "sha256": "47ea2db78b80b790758986fab8f9c98be79e80554f6cbd81ce0df9e3fe6217c6",
        "transcript": "trece",
        "sourceFormat": "pcm16",
        "durationSeconds": 0.85
      },
      "judgment": {
        "probabilityCorrect": 0.2298,
        "isCorrect": false,
        "elapsedMs": 222,
        "judgedAt": "2026-11-06T22:24:57.146Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “thirty” in Spanish.\nExpected response: treinta\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.2298
            }
          },
          "usage": {
            "input_tokens": 217,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "treinta-control",
      "category": "Correct control",
      "title": "Thirty, correctly",
      "prompt": "Say “thirty” in Spanish.",
      "expectedText": "treinta",
      "spokenText": "treinta",
      "expectedTranslation": "thirty",
      "spokenTranslation": "thirty",
      "expectedCorrect": true,
      "note": "The response supplies the requested number, thirty.",
      "audio": {
        "src": "/audio/treinta-control.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:24:58.462Z",
        "sha256": "2d9c14dfb73c01d525e597106fb9ffb83f8e0c7c171de4d79a85b15c58ff99be",
        "transcript": "treinta",
        "sourceFormat": "pcm16",
        "durationSeconds": 0.95
      },
      "judgment": {
        "probabilityCorrect": 0.9813,
        "isCorrect": true,
        "elapsedMs": 261,
        "judgedAt": "2026-11-06T22:24:58.724Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “thirty” in Spanish.\nExpected response: treinta\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9813
            }
          },
          "usage": {
            "input_tokens": 218,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "negation-control",
      "category": "Correct control",
      "title": "No coffee, as requested",
      "prompt": "Say “I don't want coffee” in Spanish.",
      "expectedText": "No quiero café.",
      "spokenText": "No quiero café.",
      "expectedTranslation": "I don't want coffee.",
      "spokenTranslation": "I don't want coffee.",
      "expectedCorrect": true,
      "note": "The response preserves the negation and correctly declines coffee.",
      "audio": {
        "src": "/audio/negation-control.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:25:00.417Z",
        "sha256": "3d4e343ac5b2763a3745be52580e4f830faf67f8c2bc0da6311225389a9a08bd",
        "transcript": "No quiero café.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.5
      },
      "judgment": {
        "probabilityCorrect": 0.9884,
        "isCorrect": true,
        "elapsedMs": 235,
        "judgedAt": "2026-11-06T22:25:00.653Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “I don't want coffee” in Spanish.\nExpected response: No quiero café.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9884
            }
          },
          "usage": {
            "input_tokens": 230,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "past-future",
      "category": "Changed tense",
      "title": "Went or going?",
      "prompt": "Say “I went to the market” in Spanish.",
      "expectedText": "Fui al mercado.",
      "spokenText": "Voy al mercado.",
      "expectedTranslation": "I went to the market.",
      "spokenTranslation": "I am going to the market.",
      "expectedCorrect": false,
      "note": "Voy describes going now or in the near future, not a completed trip in the past.",
      "audio": {
        "src": "/audio/past-future.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:25:01.858Z",
        "sha256": "462f38b2c135921cbaa0eca921df3039eb3c1b0dcfaf6c1096c750d8cb4aa70f",
        "transcript": "Voy al mercado.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.85
      },
      "judgment": {
        "probabilityCorrect": 0.8973,
        "isCorrect": true,
        "elapsedMs": 277,
        "judgedAt": "2026-11-06T22:25:02.136Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “I went to the market” in Spanish.\nExpected response: Fui al mercado.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.8973
            }
          },
          "usage": {
            "input_tokens": 235,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "past-tense-control",
      "category": "Correct control",
      "title": "A completed trip",
      "prompt": "Say “I went to the market” in Spanish.",
      "expectedText": "Fui al mercado.",
      "spokenText": "Fui al mercado.",
      "expectedTranslation": "I went to the market.",
      "spokenTranslation": "I went to the market.",
      "expectedCorrect": true,
      "note": "Fui expresses the completed past action requested by the drill.",
      "audio": {
        "src": "/audio/past-tense-control.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:25:03.431Z",
        "sha256": "4d9801c937a4354eb1a62724e467a8c41be1b6eed6cb7f6542675d44e66bb3a4",
        "transcript": "Fui al mercado.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.25
      },
      "judgment": {
        "probabilityCorrect": 0.989,
        "isCorrect": true,
        "elapsedMs": 271,
        "judgedAt": "2026-11-06T22:25:03.703Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “I went to the market” in Spanish.\nExpected response: Fui al mercado.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.989
            }
          },
          "usage": {
            "input_tokens": 228,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "left-right",
      "category": "Opposite direction",
      "title": "Left or right?",
      "prompt": "Say “Turn right” in Spanish.",
      "expectedText": "Gira a la derecha.",
      "spokenText": "Gira a la izquierda.",
      "expectedTranslation": "Turn right.",
      "spokenTranslation": "Turn left.",
      "expectedCorrect": false,
      "note": "Izquierda directs the listener left, the opposite of the requested derecha (right).",
      "audio": {
        "src": "/audio/left-right.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:25:05.049Z",
        "sha256": "976a91295dd6b538c01d2a3a4c563f505ae96038171285a8d051886341a10b11",
        "transcript": "Gira a la izquierda.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.55
      },
      "judgment": {
        "probabilityCorrect": 0.0127,
        "isCorrect": false,
        "elapsedMs": 238,
        "judgedAt": "2026-11-06T22:25:05.287Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “Turn right” in Spanish.\nExpected response: Gira a la derecha.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.0127
            }
          },
          "usage": {
            "input_tokens": 230,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "direction-control",
      "category": "Correct control",
      "title": "The right direction",
      "prompt": "Say “Turn right” in Spanish.",
      "expectedText": "Gira a la derecha.",
      "spokenText": "Gira a la derecha.",
      "expectedTranslation": "Turn right.",
      "spokenTranslation": "Turn right.",
      "expectedCorrect": true,
      "note": "The recording gives the requested right-turn instruction.",
      "audio": {
        "src": "/audio/direction-control.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:25:06.550Z",
        "sha256": "0231641e5573fc83a4b05f42c6cf271f290038a2be6fecb3802dba79df431861",
        "transcript": "Gira a la derecha.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.35
      },
      "judgment": {
        "probabilityCorrect": 0.986,
        "isCorrect": true,
        "elapsedMs": 199,
        "judgedAt": "2026-11-06T22:25:06.750Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “Turn right” in Spanish.\nExpected response: Gira a la derecha.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.986
            }
          },
          "usage": {
            "input_tokens": 228,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "two-three-tickets",
      "category": "Changed number",
      "title": "Two tickets or three?",
      "prompt": "Say “I want two tickets” in Spanish.",
      "expectedText": "Quiero dos entradas.",
      "spokenText": "Quiero tres entradas.",
      "expectedTranslation": "I want two tickets.",
      "spokenTranslation": "I want three tickets.",
      "expectedCorrect": false,
      "note": "The response changes the requested quantity from two to three.",
      "audio": {
        "src": "/audio/two-three-tickets.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:25:07.947Z",
        "sha256": "5b14fc1eac256997fd815749f828df7fde9c3cc498512b34b44e92f6c0906bac",
        "transcript": "Quiero tres entradas.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.65
      },
      "judgment": {
        "probabilityCorrect": 0.0132,
        "isCorrect": false,
        "elapsedMs": 388,
        "judgedAt": "2026-11-06T22:25:08.336Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “I want two tickets” in Spanish.\nExpected response: Quiero dos entradas.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.0132
            }
          },
          "usage": {
            "input_tokens": 233,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "plural-control",
      "category": "Correct control",
      "title": "Sleeping cats",
      "prompt": "Say “The cats sleep” in Spanish.",
      "expectedText": "Los gatos duermen.",
      "spokenText": "Los gatos duermen.",
      "expectedTranslation": "The cats sleep.",
      "spokenTranslation": "The cats sleep.",
      "expectedCorrect": true,
      "note": "The plural noun and verb correctly express the requested sentence.",
      "audio": {
        "src": "/audio/plural-control.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:25:09.581Z",
        "sha256": "e83266ace95ea50ac7da8b025d3798e7f2c3c4e7efbe182b80f222343e3ba1ee",
        "transcript": "Los gatos duermen.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.6
      },
      "judgment": {
        "probabilityCorrect": 0.9868,
        "isCorrect": true,
        "elapsedMs": 192,
        "judgedAt": "2026-11-06T22:25:09.773Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “The cats sleep” in Spanish.\nExpected response: Los gatos duermen.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9868
            }
          },
          "usage": {
            "input_tokens": 231,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "greeting-control",
      "category": "Correct control",
      "title": "A good morning",
      "prompt": "Say “Good morning” in Spanish.",
      "expectedText": "Buenos días.",
      "spokenText": "Buenos días.",
      "expectedTranslation": "Good morning.",
      "spokenTranslation": "Good morning.",
      "expectedCorrect": true,
      "note": "The recording provides the standard Spanish morning greeting.",
      "audio": {
        "src": "/audio/greeting-control.mp3",
        "model": "openai/gpt-audio-mini",
        "voice": "alloy",
        "generatedAt": "2026-11-06T22:25:11.227Z",
        "sha256": "42325eca16ae50e09f438f44370b4708c8ccd32c577c043cfeffa7668f062919",
        "transcript": "Buenos días.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.1
      },
      "judgment": {
        "probabilityCorrect": 0.9884,
        "isCorrect": true,
        "elapsedMs": 267,
        "judgedAt": "2026-11-06T22:25:11.494Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “Good morning” in Spanish.\nExpected response: Buenos días.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9884
            }
          },
          "usage": {
            "input_tokens": 221,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "category": "Similar sounds",
      "title": "Beach or silver?",
      "prompt": "Say “beach” in Spanish.",
      "expectedText": "playa",
      "spokenText": "plata",
      "expectedTranslation": "beach",
      "spokenTranslation": "silver",
      "expectedCorrect": false,
      "note": "Plata means “silver,” not “beach” (playa); a consonant changes the word.",
      "id": "playa-plata-clean",
      "audio": {
        "src": "/audio/playa-plata-clean.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-06T23:46:10.731Z",
        "sha256": "0b98171f880bbbfdea59fb9a56fab92336e1fc7b2c631c6d819ceb1be91e12d8",
        "transcript": "plata",
        "sourceFormat": "pcm16",
        "durationSeconds": 4.35
      },
      "judgment": {
        "probabilityCorrect": 0.0527,
        "isCorrect": false,
        "elapsedMs": 350,
        "judgedAt": "2026-11-06T23:48:08.773Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “beach” in Spanish.\nExpected response: playa\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.0527
            }
          },
          "usage": {
            "input_tokens": 262,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "category": "Changed meaning",
      "title": "Hungry or thirsty?",
      "prompt": "Say “I am hungry” in Spanish.",
      "expectedText": "Tengo hambre.",
      "spokenText": "Tengo sed.",
      "expectedTranslation": "I am hungry.",
      "spokenTranslation": "I am thirsty.",
      "expectedCorrect": false,
      "note": "Tengo sed expresses thirst, not the requested hunger (tengo hambre).",
      "id": "hunger-thirst-clean",
      "audio": {
        "src": "/audio/hunger-thirst-clean.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-06T23:46:11.860Z",
        "sha256": "3dec9148daf747d4798f78cec3b288618aa11dcb95a6bbee07b89429b09d2d1b",
        "transcript": "Tengo sed.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.25
      },
      "judgment": {
        "probabilityCorrect": 0.0202,
        "isCorrect": false,
        "elapsedMs": 166,
        "judgedAt": "2026-11-06T23:48:09.131Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “I am hungry” in Spanish.\nExpected response: Tengo hambre.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.0202
            }
          },
          "usage": {
            "input_tokens": 226,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "category": "Changed number",
      "title": "Forty or fourteen?",
      "prompt": "Say “forty” in Spanish.",
      "expectedText": "cuarenta",
      "spokenText": "catorce",
      "expectedTranslation": "forty",
      "spokenTranslation": "fourteen",
      "expectedCorrect": false,
      "note": "Catorce is fourteen; the requested number is forty (cuarenta).",
      "id": "forty-fourteen-clean",
      "audio": {
        "src": "/audio/forty-fourteen-clean.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-06T23:46:13.447Z",
        "sha256": "c2f00f5e73be5e635bbce2e896ae6b7303ba4fe8972038856acf677622150074",
        "transcript": "catorce",
        "sourceFormat": "pcm16",
        "durationSeconds": 3.25
      },
      "judgment": {
        "probabilityCorrect": 0.0208,
        "isCorrect": false,
        "elapsedMs": 196,
        "judgedAt": "2026-11-06T23:48:09.541Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “forty” in Spanish.\nExpected response: cuarenta\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.0208
            }
          },
          "usage": {
            "input_tokens": 250,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "category": "Changed meaning",
      "title": "Tomorrow or today?",
      "prompt": "Say “See you tomorrow” in Spanish.",
      "expectedText": "Nos vemos mañana.",
      "spokenText": "Nos vemos hoy.",
      "expectedTranslation": "See you tomorrow.",
      "spokenTranslation": "See you today.",
      "expectedCorrect": false,
      "note": "Hoy changes the meeting time to today rather than tomorrow (mañana).",
      "id": "tomorrow-today-clean",
      "audio": {
        "src": "/audio/tomorrow-today-clean.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-06T23:46:14.859Z",
        "sha256": "cef42d3ddaa6a4cb95a0b79e8efdc5e3d275087a93c17014c5318edec8c5e5db",
        "transcript": "Nos vemos hoy.",
        "sourceFormat": "pcm16",
        "durationSeconds": 2.1
      },
      "judgment": {
        "probabilityCorrect": 0.0118,
        "isCorrect": false,
        "elapsedMs": 158,
        "judgedAt": "2026-11-06T23:48:09.926Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “See you tomorrow” in Spanish.\nExpected response: Nos vemos mañana.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.0118
            }
          },
          "usage": {
            "input_tokens": 237,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "category": "Opposite meaning",
      "title": "Hot or cold?",
      "prompt": "Say “It is hot” in Spanish, describing the temperature of an object.",
      "expectedText": "Está caliente.",
      "spokenText": "Está frío.",
      "expectedTranslation": "It is hot.",
      "spokenTranslation": "It is cold.",
      "expectedCorrect": false,
      "note": "Está frío describes a cold object, the opposite of the requested está caliente.",
      "id": "hot-cold-clean",
      "audio": {
        "src": "/audio/hot-cold-clean.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-06T23:46:16.531Z",
        "sha256": "dfd91fb46828b2c09522ae3678659023e9b658ae7459d5b26aedef44d3ac356e",
        "transcript": "Está frío.",
        "sourceFormat": "pcm16",
        "durationSeconds": 4.45
      },
      "judgment": {
        "probabilityCorrect": 0.0128,
        "isCorrect": false,
        "elapsedMs": 432,
        "judgedAt": "2026-11-06T23:48:10.567Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “It is hot” in Spanish, describing the temperature of an object.\nExpected response: Está caliente.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.0128
            }
          },
          "usage": {
            "input_tokens": 272,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "category": "Correct control",
      "title": "Water, correctly",
      "prompt": "Say “water” in Spanish.",
      "expectedText": "agua",
      "spokenText": "agua",
      "expectedTranslation": "water",
      "spokenTranslation": "water",
      "expectedCorrect": true,
      "note": "Agua is the requested Spanish word for water.",
      "id": "water-control-clean",
      "audio": {
        "src": "/audio/water-control-clean.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-06T23:46:17.730Z",
        "sha256": "a4dbb0f98e66a1328fe3a726e60ef26e1fabc814ddb3686caa61d7fb79ff95bd",
        "transcript": "agua",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.45
      },
      "judgment": {
        "probabilityCorrect": 0.9879,
        "isCorrect": true,
        "elapsedMs": 246,
        "judgedAt": "2026-11-06T23:48:11.061Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “water” in Spanish.\nExpected response: agua\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9879
            }
          },
          "usage": {
            "input_tokens": 223,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "category": "Correct control",
      "title": "A nearby station",
      "prompt": "Say “The station is nearby” in Spanish.",
      "expectedText": "La estación está cerca.",
      "spokenText": "La estación está cerca.",
      "expectedTranslation": "The station is nearby.",
      "spokenTranslation": "The station is nearby.",
      "expectedCorrect": true,
      "note": "The response correctly says that the station is nearby.",
      "id": "station-control-clean",
      "audio": {
        "src": "/audio/station-control-clean.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-06T23:46:19.372Z",
        "sha256": "7d3881b889826427a0223e14fd80ab75c6367cae69bf4b3a6db35eefe2486eb8",
        "transcript": "La estación está cerca.",
        "sourceFormat": "pcm16",
        "durationSeconds": 4.15
      },
      "judgment": {
        "probabilityCorrect": 0.984,
        "isCorrect": true,
        "elapsedMs": 174,
        "judgedAt": "2026-11-06T23:48:11.463Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “The station is nearby” in Spanish.\nExpected response: La estación está cerca.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.984
            }
          },
          "usage": {
            "input_tokens": 265,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "category": "Correct control",
      "title": "No meat, as requested",
      "prompt": "Say “I do not eat meat” in Spanish.",
      "expectedText": "No como carne.",
      "spokenText": "No como carne.",
      "expectedTranslation": "I do not eat meat.",
      "spokenTranslation": "I do not eat meat.",
      "expectedCorrect": true,
      "note": "No preserves the requested negation: the speaker does not eat meat.",
      "id": "meat-negation-control-clean",
      "audio": {
        "src": "/audio/meat-negation-control-clean.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-06T23:46:20.535Z",
        "sha256": "cd64bfad5c02e9bd1712ac95cdb78ea85da9457a7a595e07ce5b5cac0e922024",
        "transcript": "No como carne.",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.45
      },
      "judgment": {
        "probabilityCorrect": 0.9882,
        "isCorrect": true,
        "elapsedMs": 206,
        "judgedAt": "2026-11-06T23:48:11.831Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “I do not eat meat” in Spanish.\nExpected response: No como carne.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9882
            }
          },
          "usage": {
            "input_tokens": 229,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "category": "Correct control",
      "title": "A twenty-euro price",
      "prompt": "Say “It costs twenty euros” in Spanish.",
      "expectedText": "Cuesta veinte euros.",
      "spokenText": "Cuesta veinte euros.",
      "expectedTranslation": "It costs twenty euros.",
      "spokenTranslation": "It costs twenty euros.",
      "expectedCorrect": true,
      "note": "The response gives the requested price of twenty euros.",
      "id": "twenty-euros-control-clean",
      "audio": {
        "src": "/audio/twenty-euros-control-clean.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-06T23:46:22.217Z",
        "sha256": "84df800e32964a1d1cf3c1ffb546466a969954e081a2978267f5e319947de75c",
        "transcript": "Cuesta veinte euros.",
        "sourceFormat": "pcm16",
        "durationSeconds": 4
      },
      "judgment": {
        "probabilityCorrect": 0.9876,
        "isCorrect": true,
        "elapsedMs": 205,
        "judgedAt": "2026-11-06T23:48:12.255Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “It costs twenty euros” in Spanish.\nExpected response: Cuesta veinte euros.\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9876
            }
          },
          "usage": {
            "input_tokens": 263,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "category": "Correct control",
      "title": "Finding the bathroom",
      "prompt": "Say “Where is the bathroom?” in Spanish.",
      "expectedText": "¿Dónde está el baño?",
      "spokenText": "¿Dónde está el baño?",
      "expectedTranslation": "Where is the bathroom?",
      "spokenTranslation": "Where is the bathroom?",
      "expectedCorrect": true,
      "note": "The response correctly asks where the bathroom is.",
      "id": "bathroom-control-clean",
      "audio": {
        "src": "/audio/bathroom-control-clean.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-06T23:46:23.489Z",
        "sha256": "e721db761a429d709f43a2d5af588c1466b5e6c0b671dcda29e2fa3c2d4e2bc7",
        "transcript": "¿Dónde está el baño?",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.85
      },
      "judgment": {
        "probabilityCorrect": 0.989,
        "isCorrect": true,
        "elapsedMs": 206,
        "judgedAt": "2026-11-06T23:48:12.641Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “Where is the bathroom?” in Spanish.\nExpected response: ¿Dónde está el baño?\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.989
            }
          },
          "usage": {
            "input_tokens": 237,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "peso-beso",
      "category": "Similar sounds",
      "title": "Weight or kiss?",
      "prompt": "Say “weight” in Spanish.",
      "expectedText": "peso",
      "spokenText": "beso",
      "expectedTranslation": "weight",
      "spokenTranslation": "kiss",
      "expectedCorrect": false,
      "note": "Changing the initial consonant from p to b changes peso (weight) into beso (kiss).",
      "audio": {
        "src": "/audio/peso-beso.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:06.759Z",
        "sha256": "2bdc61fc05cec2fa62ee3617ca20ec102905b337e463e3c9a0738938d3e68445",
        "transcript": "beso",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.5
      },
      "judgment": {
        "probabilityCorrect": 0.9065,
        "isCorrect": true,
        "elapsedMs": 227,
        "judgedAt": "2026-11-07T00:03:37.396Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “weight” in Spanish.\nExpected response: peso\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9065
            }
          },
          "usage": {
            "input_tokens": 224,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "pata-bata",
      "category": "Similar sounds",
      "title": "An animal’s leg or robe?",
      "prompt": "Say “an animal’s leg” in Spanish, using “pata”.",
      "expectedText": "pata",
      "spokenText": "bata",
      "expectedTranslation": "an animal’s leg",
      "spokenTranslation": "robe",
      "expectedCorrect": false,
      "note": "The initial consonant distinguishes pata (an animal’s leg) from bata (robe).",
      "audio": {
        "src": "/audio/pata-bata.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:08.272Z",
        "sha256": "1338b626eec0fd11c4ae066e41d72e83a38bbfe065ab153fb638d3881ac099a4",
        "transcript": "bata",
        "sourceFormat": "pcm16",
        "durationSeconds": 4.25
      },
      "judgment": {
        "probabilityCorrect": 0.9731,
        "isCorrect": true,
        "elapsedMs": 194,
        "judgedAt": "2026-11-07T00:03:37.593Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “an animal’s leg” in Spanish, using “pata”.\nExpected response: pata\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9731
            }
          },
          "usage": {
            "input_tokens": 270,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "pala-bala",
      "category": "Similar sounds",
      "title": "Shovel or bullet?",
      "prompt": "Say “shovel” in Spanish.",
      "expectedText": "pala",
      "spokenText": "bala",
      "expectedTranslation": "shovel",
      "spokenTranslation": "bullet",
      "expectedCorrect": false,
      "note": "An initial b instead of p changes pala (shovel) into bala (bullet).",
      "audio": {
        "src": "/audio/pala-bala.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:09.753Z",
        "sha256": "e1c99063b31f9fde8c69991961f3da3bdb231508fcd4d075ffc1212c66bdd3dc",
        "transcript": "bala",
        "sourceFormat": "pcm16",
        "durationSeconds": 3.15
      },
      "judgment": {
        "probabilityCorrect": 0.5108,
        "isCorrect": true,
        "elapsedMs": 321,
        "judgedAt": "2026-11-07T00:03:37.916Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “shovel” in Spanish.\nExpected response: pala\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.5108
            }
          },
          "usage": {
            "input_tokens": 247,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "jamon-jabon",
      "category": "Similar sounds",
      "title": "Ham or soap?",
      "prompt": "Say “ham” in Spanish.",
      "expectedText": "jamón",
      "spokenText": "jabón",
      "expectedTranslation": "ham",
      "spokenTranslation": "soap",
      "expectedCorrect": false,
      "note": "The middle consonant changes jamón (ham) into jabón (soap); both stress the final syllable.",
      "audio": {
        "src": "/audio/jamon-jabon.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:10.843Z",
        "sha256": "9e88821e29080c3e45f7fb66e536f2c8cf98c4f72d252960ccd82fe7a572091e",
        "transcript": "jabón",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.3
      },
      "judgment": {
        "probabilityCorrect": 0.5807,
        "isCorrect": true,
        "elapsedMs": 377,
        "judgedAt": "2026-11-07T00:03:38.294Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “ham” in Spanish.\nExpected response: jamón\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.5807
            }
          },
          "usage": {
            "input_tokens": 222,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "pino-vino",
      "category": "Similar sounds",
      "title": "Pine tree or wine?",
      "prompt": "Say “pine tree” in Spanish.",
      "expectedText": "pino",
      "spokenText": "vino",
      "expectedTranslation": "pine tree",
      "spokenTranslation": "wine",
      "expectedCorrect": false,
      "note": "The initial p in pino (pine tree) differs from the b sound written v in vino (wine).",
      "audio": {
        "src": "/audio/pino-vino.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:12.351Z",
        "sha256": "7cd58cc2c76ca4fe20adc08d2f0c8753fc487f47f213a350c80f777edd636a58",
        "transcript": "vino",
        "sourceFormat": "pcm16",
        "durationSeconds": 2.25
      },
      "judgment": {
        "probabilityCorrect": 0.963,
        "isCorrect": true,
        "elapsedMs": 255,
        "judgedAt": "2026-11-07T00:03:38.552Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “pine tree” in Spanish.\nExpected response: pino\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.963
            }
          },
          "usage": {
            "input_tokens": 236,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "pelo-pero",
      "category": "Similar sounds",
      "title": "Hair or but?",
      "prompt": "Say “hair” in Spanish.",
      "expectedText": "pelo",
      "spokenText": "pero",
      "expectedTranslation": "hair",
      "spokenTranslation": "but",
      "expectedCorrect": false,
      "note": "Replacing l with a tapped r changes pelo (hair) into pero (but).",
      "audio": {
        "src": "/audio/pelo-pero.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:13.762Z",
        "sha256": "27a16619367e3d518a00316f73837a49f096aa26373f62ddef774563ca13d3b9",
        "transcript": "pero",
        "sourceFormat": "pcm16",
        "durationSeconds": 3.65
      },
      "judgment": {
        "probabilityCorrect": 0.7131,
        "isCorrect": true,
        "elapsedMs": 207,
        "judgedAt": "2026-11-07T00:03:38.761Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “hair” in Spanish.\nExpected response: pelo\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.7131
            }
          },
          "usage": {
            "input_tokens": 252,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "loma-lona",
      "category": "Similar sounds",
      "title": "Hill or canvas?",
      "prompt": "Say “hill” in Spanish, using “loma”.",
      "expectedText": "loma",
      "spokenText": "lona",
      "expectedTranslation": "hill",
      "spokenTranslation": "canvas",
      "expectedCorrect": false,
      "note": "The nasal consonant changes loma (hill) into lona (canvas): m becomes n.",
      "audio": {
        "src": "/audio/loma-lona.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:19.323Z",
        "sha256": "e186dc097dc49ddcb32783796324861b47a06c9a4425536e09a098d300a2e202",
        "transcript": "lona",
        "sourceFormat": "pcm16",
        "durationSeconds": 2.5
      },
      "judgment": {
        "probabilityCorrect": 0.8989,
        "isCorrect": true,
        "elapsedMs": 189,
        "judgedAt": "2026-11-07T00:03:39.778Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “hill” in Spanish, using “loma”.\nExpected response: loma\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.8989
            }
          },
          "usage": {
            "input_tokens": 244,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "lago-lado",
      "category": "Similar sounds",
      "title": "Lake or side?",
      "prompt": "Say “lake” in Spanish.",
      "expectedText": "lago",
      "spokenText": "lado",
      "expectedTranslation": "lake",
      "spokenTranslation": "side",
      "expectedCorrect": false,
      "note": "The middle consonant distinguishes lago (lake) from lado (side).",
      "audio": {
        "src": "/audio/lago-lado.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:20.660Z",
        "sha256": "fb7c90f2e44c4f0e48a884ff225562a7b082347edab2c2c0961ed2f07b8a697e",
        "transcript": "lado",
        "sourceFormat": "pcm16",
        "durationSeconds": 2.25
      },
      "judgment": {
        "probabilityCorrect": 0.6292,
        "isCorrect": true,
        "elapsedMs": 333,
        "judgedAt": "2026-11-07T00:03:40.112Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “lake” in Spanish.\nExpected response: lago\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.6292
            }
          },
          "usage": {
            "input_tokens": 235,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "manga-mango",
      "category": "Similar sounds",
      "title": "Sleeve or mango?",
      "prompt": "Say “sleeve” in Spanish.",
      "expectedText": "manga",
      "spokenText": "mango",
      "expectedTranslation": "sleeve",
      "spokenTranslation": "mango",
      "expectedCorrect": false,
      "note": "Changing the final vowel from a to o changes manga (sleeve) into mango (the fruit).",
      "audio": {
        "src": "/audio/manga-mango.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:22.142Z",
        "sha256": "fc777db092e0d7e9d6569f4745f31db928f5fdd3ebde699159d7f0c16a5f101e",
        "transcript": "mango",
        "sourceFormat": "pcm16",
        "durationSeconds": 3.85
      },
      "judgment": {
        "probabilityCorrect": 0.9724,
        "isCorrect": true,
        "elapsedMs": 200,
        "judgedAt": "2026-11-07T00:03:40.313Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “sleeve” in Spanish.\nExpected response: manga\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9724
            }
          },
          "usage": {
            "input_tokens": 256,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "boca-poca",
      "category": "Similar sounds",
      "title": "Mouth or little?",
      "prompt": "Say “mouth” in Spanish.",
      "expectedText": "boca",
      "spokenText": "poca",
      "expectedTranslation": "mouth",
      "spokenTranslation": "little (feminine)",
      "expectedCorrect": false,
      "note": "The initial consonant changes boca (mouth) into poca, the feminine form of “little.”",
      "audio": {
        "src": "/audio/boca-poca.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:24.866Z",
        "sha256": "5d6bda1b19b602c8e94e91fb7b1ee97d79f6fcec916b58306fd6c391b1369e7b",
        "transcript": "poca",
        "sourceFormat": "pcm16",
        "durationSeconds": 3.05
      },
      "judgment": {
        "probabilityCorrect": 0.9877,
        "isCorrect": true,
        "elapsedMs": 283,
        "judgedAt": "2026-11-07T00:03:40.597Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “mouth” in Spanish.\nExpected response: boca\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9877
            }
          },
          "usage": {
            "input_tokens": 244,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "hambre-hombre",
      "category": "Similar sounds",
      "title": "Hunger or man?",
      "prompt": "Say “hunger” in Spanish.",
      "expectedText": "hambre",
      "spokenText": "hombre",
      "expectedTranslation": "hunger",
      "spokenTranslation": "man",
      "expectedCorrect": false,
      "note": "The first vowel distinguishes hambre (hunger) from hombre (man).",
      "audio": {
        "src": "/audio/hambre-hombre.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:27.189Z",
        "sha256": "80068b5d00885abbb747dc4f00527194a22a3dea6b5870ba1fbf3c6d20479791",
        "transcript": "hombre",
        "sourceFormat": "pcm16",
        "durationSeconds": 3.75
      },
      "judgment": {
        "probabilityCorrect": 0.9421,
        "isCorrect": true,
        "elapsedMs": 207,
        "judgedAt": "2026-11-07T00:03:40.805Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “hunger” in Spanish.\nExpected response: hambre\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.9421
            }
          },
          "usage": {
            "input_tokens": 255,
            "output_tokens": 0
          }
        }
      }
    },
    {
      "id": "oveja-abeja",
      "category": "Similar sounds",
      "title": "Sheep or bee?",
      "prompt": "Say “sheep” in Spanish.",
      "expectedText": "oveja",
      "spokenText": "abeja",
      "expectedTranslation": "sheep",
      "spokenTranslation": "bee",
      "expectedCorrect": false,
      "note": "The first vowel changes oveja (sheep) into abeja (bee); Spanish b and v represent the same consonant phoneme.",
      "audio": {
        "src": "/audio/oveja-abeja.mp3",
        "model": "openai/gpt-audio",
        "voice": "alloy",
        "generatedAt": "2026-11-07T00:01:28.344Z",
        "sha256": "5066cc02a6366b292e0e6869e6af780a247df9966012b1c216846b66c27b2da6",
        "transcript": "abeja",
        "sourceFormat": "pcm16",
        "durationSeconds": 1.45
      },
      "judgment": {
        "probabilityCorrect": 0.1485,
        "isCorrect": false,
        "elapsedMs": 203,
        "judgedAt": "2026-11-07T00:03:41.010Z",
        "model": "@cf/cloudflare/clef-omni",
        "request": {
          "model": "clef-omni",
          "state": "This is a Spanish language drill. Drill: Say “sheep” in Spanish.\nExpected response: oveja\nGrade correctness from the actual recording, not from the expected response alone.",
          "questions": {
            "is_correct": {
              "type": "noul",
              "instructions": "Does the spoken Spanish response correctly answer the drill? Judge the actual audio. Reject different words, changed numbers, or missing negation; tolerate natural accent variation."
            }
          }
        },
        "response": {
          "model": "clef-omni",
          "answers": {
            "is_correct": {
              "type": "noul",
              "noul": 0.1485
            }
          },
          "usage": {
            "input_tokens": 226,
            "output_tokens": 0
          }
        }
      }
    }
  ]
}