ground-zero / configs /eval /phrasebook_eval.json
ous-sow
Claude Opus 5.5
Phrasebook eval harness + accent/filler-tolerant matching
ea5dc26
Raw History Blame Contribute Delete
8.52 kB
{
"notes": "Held-out phrasebook eval. Inputs are written to differ from the curated aliases (typos, casing, code-switching, extra words). `expect` is the canonical English `source` of the row that should match, or null when the input should NOT short-circuit (open questions the LLM must handle). Langs: both = run against bam and ful.",
"items": [
{
"q": "good mornin",
"lang": "both",
"expect": "Good morning"
},
{
"q": "Good morning!!",
"lang": "both",
"expect": "Good morning"
},
{
"q": "bonjour à vous",
"lang": "both",
"expect": "Hello / Thank you"
},
{
"q": "bonsoir",
"lang": "both",
"expect": "Good evening"
},
{
"q": "how r you",
"lang": "both",
"expect": "How are you?"
},
{
"q": "ca va?",
"lang": "both",
"expect": "How are you?"
},
{
"q": "comment allez vous",
"lang": "both",
"expect": "How are you?"
},
{
"q": "thank you so much",
"lang": "both",
"expect": "Hello / Thank you"
},
{
"q": "merci beaucoup",
"lang": "both",
"expect": "Hello / Thank you"
},
{
"q": "whats your name",
"lang": "both",
"expect": "What is your name?"
},
{
"q": "comment tu t appelles",
"lang": "both",
"expect": "What is your name?"
},
{
"q": "my name is Awa",
"lang": "both",
"expect": "My name is..."
},
{
"q": "how is ur family",
"lang": "both",
"expect": "How is the family?"
},
{
"q": "la famille va bien ?",
"lang": "both",
"expect": "How is the family?"
},
{
"q": "how are the children doing",
"lang": "both",
"expect": "How are the children?"
},
{
"q": "Im hungry",
"lang": "both",
"expect": "I am hungry"
},
{
"q": "j ai soif",
"lang": "both",
"expect": "I am thirsty"
},
{
"q": "give me some water",
"lang": "both",
"expect": "Give me water"
},
{
"q": "rice",
"lang": "both",
"expect": "Rice"
},
{
"q": "du riz",
"lang": "both",
"expect": "Rice"
},
{
"q": "sugar please",
"lang": "both",
"expect": "Sugar"
},
{
"q": "how is the farming going",
"lang": "both",
"expect": "How is the farming?"
},
{
"q": "comment va le champ",
"lang": "both",
"expect": "How is the farming?"
},
{
"q": "the field",
"lang": "both",
"expect": "The field"
},
{
"q": "sheep",
"lang": "both",
"expect": "Sheep"
},
{
"q": "une chèvre",
"lang": "both",
"expect": "Goat"
},
{
"q": "Im sick",
"lang": "both",
"expect": "I am sick"
},
{
"q": "je suis malade",
"lang": "both",
"expect": "I am sick"
},
{
"q": "j'ai mal a la tete",
"lang": "both",
"expect": "My head hurts"
},
{
"q": "my stomach is hurting",
"lang": "both",
"expect": "My stomach hurts"
},
{
"q": "i have fever",
"lang": "both",
"expect": "I have fever"
},
{
"q": "where is the doctor",
"lang": "both",
"expect": "Where is the doctor?"
},
{
"q": "drink this",
"lang": "both",
"expect": "Drink this"
},
{
"q": "how much is it?",
"lang": "both",
"expect": "@price"
},
{
"q": "c'est combien",
"lang": "both",
"expect": "@price"
},
{
"q": "too expensive",
"lang": "both",
"expect": "@expensive"
},
{
"q": "i dont have money",
"lang": "both",
"expect": "I have no money"
},
{
"q": "market",
"lang": "both",
"expect": "Market"
},
{
"q": "soap",
"lang": "both",
"expect": "Soap"
},
{
"q": "salt",
"lang": "both",
"expect": "Salt"
},
{
"q": "is it far",
"lang": "both",
"expect": "Is it far?"
},
{
"q": "turn right",
"lang": "both",
"expect": "Turn right"
},
{
"q": "tourne a gauche",
"lang": "both",
"expect": "Turn left"
},
{
"q": "stop here please",
"lang": "both",
"expect": "Stop here"
},
{
"q": "lets go",
"lang": "both",
"expect": "Let's go"
},
{
"q": "allons y",
"lang": "both",
"expect": "Let's go"
},
{
"q": "i understand",
"lang": "both",
"expect": "I understand"
},
{
"q": "i dont understand",
"lang": "both",
"expect": "I don't understand"
},
{
"q": "je comprends pas",
"lang": "both",
"expect": "I don't understand"
},
{
"q": "speak slowly please",
"lang": "both",
"expect": "Speak slowly"
},
{
"q": "i dont know",
"lang": "both",
"expect": "I don't know"
},
{
"q": "yes",
"lang": "both",
"expect": "Yes"
},
{
"q": "oui",
"lang": "both",
"expect": "Yes"
},
{
"q": "no",
"lang": "both",
"expect": "No"
},
{
"q": "today",
"lang": "both",
"expect": "Today"
},
{
"q": "demain",
"lang": "both",
"expect": "Tomorrow"
},
{
"q": "bye",
"lang": "both",
"expect": "Goodbye"
},
{
"q": "au revoir",
"lang": "both",
"expect": "Goodbye"
},
{
"q": "see you tomorow",
"lang": "both",
"expect": "@tomorrow_bye"
},
{
"q": "have a good day",
"lang": "both",
"expect": "Have a good day"
},
{
"q": "bonne nuit",
"lang": "both",
"expect": "Have a good night"
},
{
"q": "inchallah",
"lang": "both",
"expect": "God willing"
},
{
"q": "hamdulillah",
"lang": "both",
"expect": "Thank God"
},
{
"q": "when should I plant millet this season?",
"lang": "both",
"expect": null
},
{
"q": "my cow is not eating, what should I do",
"lang": "both",
"expect": null
},
{
"q": "how much fertilizer for one hectare",
"lang": "both",
"expect": null
},
{
"q": "will it rain this week",
"lang": "both",
"expect": null
},
{
"q": "tell me a story",
"lang": "both",
"expect": null
},
{
"q": "where can I sell my onions",
"lang": "both",
"expect": null
},
{
"q": "is the soil too dry",
"lang": "both",
"expect": null
},
{
"q": "did u sleep well",
"lang": "bam",
"expect": "How was the night?"
},
{
"q": "good job",
"lang": "bam",
"expect": "Well done"
},
{
"q": "corn",
"lang": "bam",
"expect": "Maize / Corn"
},
{
"q": "où est la daba",
"lang": "bam",
"expect": "Where is the hoe?"
},
{
"q": "do you speak bambara",
"lang": "bam",
"expect": "Do you speak Bambara?"
},
{
"q": "god bless u",
"lang": "bam",
"expect": "God bless you"
},
{
"q": "im a farmer",
"lang": "ful",
"expect": "I am a farmer"
},
{
"q": "je suis cultivateur",
"lang": "ful",
"expect": "I am a farmer"
},
{
"q": "milk",
"lang": "ful",
"expect": "Milk"
},
{
"q": "du lait",
"lang": "ful",
"expect": "Milk"
},
{
"q": "coffee",
"lang": "ful",
"expect": "Coffee"
},
{
"q": "harvest",
"lang": "ful",
"expect": "To harvest"
},
{
"q": "bike",
"lang": "ful",
"expect": "Bicycle"
},
{
"q": "do you speak french",
"lang": "ful",
"expect": "Do you speak French?"
},
{
"q": "how many kids do you have",
"lang": "ful",
"expect": "How many children?"
}
],
"groups": {
"@price": [
"How much?",
"How much is this?"
],
"@expensive": [
"It is too much",
"It is too expensive"
],
"@tomorrow_bye": [
"Until tomorrow",
"See you tomorrow"
],
"@name": [
"My name is..."
]
}
}