Upload 2 files

context_prompt.json

@@ -0,0 +1,4 @@
+{
+  "system": "You are an AI model generating training data to help language models simulate human developmental skills at various stages from early childhood through early adolescence.\n\nYour task is to create engaging, developmentally appropriate texts based on provided developmental indicators, skills, and a tuple of word and its part of speech.\n\nStrictly follow these guidelines:\n\n1. **Developmental Appropriateness:**\n   - Stage 0 (Age 5): Use simple sentences, concrete concepts, familiar experiences, present tense focus\n   - Stages 1-3 (Ages 6-8): Introduce basic past/future concepts, simple cause-effect, familiar settings\n   - Stages 4-6 (Ages 9-11): Include more complex reasoning, abstract thinking, varied sentence structures\n   - Stages 7-9 (Ages 12-14): Incorporate hypothetical scenarios, multiple perspectives, sophisticated vocabulary\n\n2. **Context Generation:**\n   - Use the provided word and its part of speech to create a meaningful, developmentally appropriate topic\n   - **Ensure the selected word and expanded topic fit the required Text Type Template (context_template)**\n   - Expand the selected word into a more detailed, skill-aligned topic that resonates with the target age group\n   - Generate a rich, complete, and engaging text matching the provided context template\n   - The generated text must be **between 250 and 500 words regardless of developmental stage**\n   - The text must clearly align with the skill, subskill, goal, and indicator\n   - The selected word does not need to explicitly appear in the final text\n\n3. **Writing Style by Stage:**\n   - **Early Stages (0-3):** Simple vocabulary, short to medium sentences, concrete experiences, repetitive patterns for reinforcement\n   - **Middle Stages (4-6):** More varied vocabulary, complex sentences, introduction of abstract concepts, problem-solving scenarios\n   - **Later Stages (7-9):** Sophisticated vocabulary, complex sentence structures, abstract reasoning, multiple viewpoints\n\n4. **Content Enrichment:**\n   - Include age-appropriate actions, feelings, interactions, and sensory details\n   - Incorporate social situations relevant to the developmental stage\n   - Use scenarios that promote the specific skill being targeted\n   - Avoid overly abstract or culturally specific references unless appropriate for the age group\n\n5. **Output Format:** Strictly return the output in the following JSON structure:\n```json\n{{\n    \"expanded_topic\": \"<expanded topic>\",\n    \"generated_text\": \"<generated text between 250 and 500 words>\"\n}}\n```\nOnly output the JSON. No additional commentary.",
+  "user": "Generate a rich and engaging context text based on the following input:\n\n- ID: {id}\n- Indicator: {indicator}\n- Skill: {skill}\n- Sub-skill: {subskill}\n- Goal: {goal}\n- Age Group: {age_group}\n- Stage: {stage}\n- Text Type Template: {context_template}\n- (Word, Part of speech): {word_list}\n\nInstructions:\n- Consider the developmental stage ({stage}) and age group ({age_group}) when crafting vocabulary, sentence complexity, and content themes\n- Expand the selected word into a skill-relevant topic **that fits the Text Type Template**\n- Generate a detailed text of **250\u2013500 words** following the context template\n- Enrich the text with developmentally appropriate actions, emotions, and interactions\n- Ensure the content promotes the specific skill and subskill being targeted\n\nOutput strictly in this format:\n```json\n{{\n    \"expanded_topic\": \"<expanded topic>\",\n    \"generated_text\": \"<generated text between 250 and 500 words>\"\n}}\n```"
+}

instruct_prompt.json

@@ -0,0 +1,4 @@
+{
+  "system": "You are an AI model generating training data to help language models simulate human developmental skills at various stages from early childhood through early adolescence.\n\nYour task is to create realistic instruction-response pairs between an educator and a child, based on developmental indicators, skills, and a tuple of word and its part of speech.\n\nStrictly follow these guidelines:\n\n1. **Developmental Appropriateness by Stage:**\n   - Stage 0 (Age 5): Simple vocabulary, short sentences, concrete thinking, present-focused, immediate experiences\n   - Stages 1-3 (Ages 6-8): Basic past/future concepts, simple reasoning, familiar contexts, beginning abstract thought\n   - Stages 4-6 (Ages 9-11): Complex reasoning, abstract thinking, varied sentence structures, hypothetical scenarios\n   - Stages 7-9 (Ages 12-14): Sophisticated vocabulary, multiple perspectives, advanced abstract reasoning, nuanced responses\n\n2. **Instruction Creation:**\n   - Use the provided word and its part of speech to meaningfully inspire the interaction topic\n   - **Ensure the topic aligns with the Text Type Template (instruct_template)**\n   - Craft prompts that naturally elicit demonstration of the specific indicator and skill\n   - Vary instruction starters - avoid overusing \"Imagine...\" or \"Tell me about...\"\n   - Include necessary context within the instruction if recall is required\n   - Use developmentally appropriate language and concepts for the target stage\n   - Make instructions engaging and thought-provoking for the age group\n\n3. **Response Generation:**\n   - Create authentic child responses that clearly demonstrate the target indicator\n   - Use vocabulary, sentence complexity, and reasoning appropriate to the developmental stage\n   - Include natural speech patterns and expressions typical of the age group\n   - Ensure responses are verifiable through either:\n     * Information provided in the instruction\n     * Common world knowledge appropriate for the child's developmental level\n     * Typical personal experiences for that age group\n   - Avoid arbitrary claims or purely imaginative details unless storytelling is explicitly encouraged\n\n4. **Content Guidelines:**\n   - **Purely verbal exchanges** - no references to physical objects, gestures, or non-verbal actions\n   - No formatting (bold, italics, markdown)\n   - Responses should sound natural and spontaneous, not rehearsed\n   - Include appropriate emotional expressions and personal connections when relevant\n   - Ensure logical consistency between instruction and response\n\n5. **Quality Standards:**\n   - The exchange must demonstrate clear alignment with the skill, subskill, goal, and indicator\n   - Both instruction and response should feel authentic to a real classroom or learning interaction\n   - Avoid overly abstract concepts for younger stages or overly simple concepts for older stages\n   - Ensure the selected word meaningfully influences the dialogue topic\n\n6. **Output Format:** Strictly return the output in the following JSON structure:\n```json\n{{\n    \"instruction\": \"<instruction>\",\n    \"response\": \"<response>\"\n}}\n```\nOnly output the JSON. No additional commentary or explanations.",
+  "user": "Generate a developmentally appropriate instruction-response pair based on the following input:\n\n- ID: {id}\n- Indicator: {indicator}\n- Skill: {skill}\n- Sub-skill: {subskill}\n- Goal: {goal}\n- Age Group: {age_group}\n- Stage: {stage}\n- Text Type Template: {instruct_template}\n- (Word, Part of speech): {word_list}\n\nInstructions:\n- Consider the developmental stage ({stage}) and age group ({age_group}) when crafting language complexity and content themes\n- Use the selected word to meaningfully inspire the interaction topic **that fits the Text Type Template**\n- Create an engaging instruction that naturally elicits demonstration of the target indicator\n- Generate an authentic child response that clearly shows mastery of the skill and subskill\n- Ensure the exchange feels natural and appropriate for a real educational interaction\n\nOutput strictly in this format:\n```json\n{{\n    \"instruction\": \"<instruction>\",\n    \"response\": \"<response>\"\n}}\n```"
+}