annotation_task_name: "Evaluation of Math Tutor Responses" task_dir: "." data_files: - "my-annotation-task/data/math_annotator_demo_all_with_id_text2show.csv" item_properties: id_key: id text_key: text2show instance_display: fields: - key: text2show_html type: html layout: direction: vertical gap: "16px" resizable: false output_annotation_dir: "annotation_output/" export_annotation_format: "json" require_password: false annotation_schemes: - annotation_type: radio required: true name: Content Correctness description: "Content Correctness: Are the facts, explanations, code, fixes, or judgments technically and factually correct?" labels: ["Yes", "To some extent", "No"] - annotation_type: radio required: true name: Learner-State Assessment description: "Learner-State Assessment: Does the tutor correctly assess whether the learner's work is correct, incorrect, incomplete, or uncertain?" labels: ["Yes", "To some extent", "No"] - annotation_type: radio required: true name: Issue Localization description: "Issue Localization: Does the tutor identify the relevant step, line, concept, or reasoning gap?" labels: ["Yes", "To some extent", "No"] - annotation_type: radio required: true name: Disclosure Appropriateness description: "Disclosure Appropriateness: Does the tutor reveal an appropriate amount of solution detail given the task type and learner state?" labels: ["Yes", "To some extent", "No"] - annotation_type: radio required: true name: Providing Guidance description: "Providing Guidance: Does the response provide useful help, such as a hint, explanation, validation, diagnostic feedback, worked step, or repair advice?" labels: ["Yes", "To some extent", "No"] - annotation_type: radio required: true name: Coherence description: "Coherence: Is the response logically consistent and generally relevant to the task?" labels: ["Yes", "To some extent", "No"] - annotation_type: radio required: true name: Actionability description: "Actionability: After reading the response, is it clear what the learner should do, think about, or check next?" labels: ["Yes", "To some extent", "No"] - annotation_type: radio required: true name: Clarity description: "Clarity: Is the response easy to understand in terms of context, structure, and references?" labels: ["Yes", "To some extent", "No"] - annotation_type: radio required: true name: Conciseness description: "Conciseness: Is the response appropriately scoped in length and detail for the context?" labels: ["Yes", "To some extent", "No"] - annotation_type: radio required: true name: Humanness description: "Humanness: Does this look or sound like what a human tutor would do or say in this context?" labels: ["Yes", "To some extent", "No"] # Generated but intentionally not enabled for the first demo: # phases: # training: # enabled: true # data_file: "my-annotation-task/data/training_questions.json" # passing_criteria: # min_correct: 8 # total_questions: 14 # retries: # enabled: true # max_retries: 2 # show_explanations: true # randomize: true # # gold_standards: # enabled: true # items_file: "my-annotation-task/data/gold_standards.json" # mode: "separate" # accuracy: # min_threshold: 0.7 # evaluation_count: 14 custom_footer_html: |