初始化项目，由ModelHub XC社区提供模型

Model: JuliaKreutzerCohere/tiny-aya-global-prompt-tasktype
Source: Original Platform

prompts/default.txt

@@ -0,0 +1,17 @@
+CONTEXT:
+{context}
+
+TASK TYPE: `{task_type}`
+
+QUERY:
+{query}
+
+Read the QUERY instructions carefully — this task type may be unfamiliar. Learn only from CONTEXT, then produce answers in exactly the form the QUERY asks for.
+
+How to solve:
+1. Deduce the linguistic rules from the CONTEXT examples only.
+2. Apply those rules to every item in QUERY, in order.
+3. Match the answer unit and format shown by the instructions and CONTEXT (words, letters, digits, transcriptions, etc.).
+4. Put only the bare required answer on each answer line — no glosses, numbering, or commentary.
+
+Draft, then verify completeness and format. Finally write a line that says exactly `FINAL ANSWERS:` and, below it, one bare answer per QUERY item in QUERY order — no numbering, no quotes, no extra text.

prompts/fill_blanks.txt

@@ -0,0 +1,17 @@
+CONTEXT:
+{context}
+
+TASK TYPE: `fill_blanks`
+
+QUERY:
+{query}
+
+This is a fill-in-the-blanks task. Missing pieces may be whole words, morphemes, or phonetic segments — match the kind of unit CONTEXT uses in the corresponding positions.
+
+How to solve:
+1. Study complete CONTEXT examples to recover the paradigm / pattern.
+2. Identify exactly what is missing in each QUERY blank (not what could be invented).
+3. Fill each blank with only the missing form, in the same style as CONTEXT (including transcription brackets if CONTEXT uses them).
+4. Return one filled form per blank, in the order blanks / items appear in QUERY.
+
+Draft, then verify every blank has an answer. Finally write a line that says exactly `FINAL ANSWERS:` and, below it, one bare filled form per blank/item — no numbering, no quotes, no extra text.

prompts/match_letters.txt

@@ -0,0 +1,17 @@
+CONTEXT:
+{context}
+
+TASK TYPE: `match_letters`
+
+QUERY:
+{query}
+
+This is a matching / multiple-choice letter task. Each QUERY item asks you to choose among labeled options (A, B, C, ...).
+
+How to solve:
+1. Extract the mapping or rule set from CONTEXT.
+2. For each numbered QUERY item, evaluate the options against that rule set.
+3. Return only the chosen option letter (A, B, C, ...), uppercase.
+4. Do not repeat the option text, explanations, or punctuation around the letter on the answer lines.
+
+Draft, then check you have one letter per QUERY item. Finally write a line that says exactly `FINAL ANSWERS:` and, below it, one bare letter per item in QUERY order — no numbering, no quotes, no extra text.

prompts/num_to_text.txt

@@ -0,0 +1,17 @@
+CONTEXT:
+{context}
+
+TASK TYPE: `num_to_text`
+
+QUERY:
+{query}
+
+This is a number-to-text task. CONTEXT shows how digit values are written as number words (or expressions) in the target language; convert each QUERY number into that written form.
+
+How to solve:
+1. Infer the number-word construction rules from CONTEXT (bases, multipliers, conjunctions, morphology).
+2. Apply them to each QUERY number.
+3. Return only the number written out in words / forms as in CONTEXT's language style.
+4. Do not output digits on the answer lines and do not add English glosses or commentary.
+
+Draft, then verify each QUERY item has a written numeral form. Finally write a line that says exactly `FINAL ANSWERS:` and, below it, one bare written form per item in QUERY order — no numbering, no quotes, no extra text.

prompts/text_to_num.txt

@@ -0,0 +1,17 @@
+CONTEXT:
+{context}
+
+TASK TYPE: `text_to_num`
+
+QUERY:
+{query}
+
+This is a text-to-number task. CONTEXT shows how number words (or number expressions) map to values; convert each QUERY expression into digits.
+
+How to solve:
+1. Infer the number system from CONTEXT (bases, place values, multipliers, word order).
+2. Apply it to each QUERY expression.
+3. Return only the numeric value in ordinary digits (e.g. 42), with no units or words.
+4. Do not spell numbers out and do not add commentary on the answer lines.
+
+Draft, then verify each QUERY item has a digit answer. Finally write a line that says exactly `FINAL ANSWERS:` and, below it, one bare number per item in QUERY order — no numbering, no quotes, no extra text.

prompts/translation.txt

@@ -0,0 +1,17 @@
+CONTEXT:
+{context}
+
+TASK TYPE: `translation`
+
+QUERY:
+{query}
+
+This is a translation task. Use only the CONTEXT examples to learn how forms map between languages (or between orthography and meaning), then translate each QUERY item.
+
+How to solve:
+1. Align CONTEXT pairs and find systematic correspondences (roots, affixes, word order, agreement).
+2. Apply those rules to each QUERY item in order.
+3. Output only the translated form itself — in the language the QUERY asks for.
+4. Do not add glosses, English explanations, punctuation wrappers, or commentary on the answer lines.
+
+Draft, then check completeness and format. Finally write a line that says exactly `FINAL ANSWERS:` and, below it, one bare translation per QUERY item, in QUERY order — no numbering, no quotes, no extra text.