Blame for project_qa_squad.jsonl

{"id": "project-001", "title": "Content Tagging and Competency Mapping", "context": "It is an AI-powered Content Tagging and Competency Mapping system that analyzes learning content and maps it to competencies from the Saudi Skills Taxonomy.", "question": "What is the purpose of the project?", "answers": {"text": ["It is an AI-powered Content Tagging and Competency Mapping system that analyzes learning content and maps it to competencies from the Saudi Skills Taxonomy."], "answer_start": [0]}}
{"id": "project-002", "title": "Content Tagging and Competency Mapping", "context": "It was developed during an SDA bootcamp with WeCloud and BeamData.", "question": "In what program was the project developed?", "answers": {"text": ["It was developed during an SDA bootcamp with WeCloud and BeamData."], "answer_start": [0]}}
{"id": "project-003", "title": "Content Tagging and Competency Mapping", "context": "It produces topic tags, proposed competencies, difficulty, confidence, notes, retrieval candidates, and chunk count.", "question": "What outputs does the system produce for learning content?", "answers": {"text": ["It produces topic tags, proposed competencies, difficulty, confidence, notes, retrieval candidates, and chunk count."], "answer_start": [0]}}
{"id": "project-004", "title": "Content Tagging and Competency Mapping", "context": "Competencies are mapped to the Saudi Skills Taxonomy.", "question": "Which taxonomy are competencies mapped to?", "answers": {"text": ["Competencies are mapped to the Saudi Skills Taxonomy."], "answer_start": [0]}}
{"id": "project-005", "title": "Content Tagging and Competency Mapping", "context": "The Saudi Skills Taxonomy contains 134 competency entries.", "question": "How many competency entries are in the Saudi Skills Taxonomy?", "answers": {"text": ["The Saudi Skills Taxonomy contains 134 competency entries."], "answer_start": [0]}}
{"id": "project-006", "title": "Content Tagging and Competency Mapping", "context": "No. 34 is the number of validated ground-truth competency mappings used in evaluation. The taxonomy contains 134 competency entries.", "question": "Is 34 the number of competencies in the taxonomy?", "answers": {"text": ["No. 34 is the number of validated ground-truth competency mappings used in evaluation. The taxonomy contains 134 competency entries."], "answer_start": [0]}}
{"id": "project-007", "title": "Content Tagging and Competency Mapping", "context": "It is the number of validated competency mappings (ground truth) used in the final evaluation.", "question": "What does the number 34 represent in the project?", "answers": {"text": ["It is the number of validated competency mappings (ground truth) used in the final evaluation."], "answer_start": [0]}}
{"id": "project-008", "title": "Content Tagging and Competency Mapping", "context": "The final evaluation used 12 official evaluation files.", "question": "How many official evaluation files were used in the final evaluation?", "answers": {"text": ["The final evaluation used 12 official evaluation files."], "answer_start": [0]}}
{"id": "project-009", "title": "Content Tagging and Competency Mapping", "context": "The taxonomy is in saudi_skills_taxonomy_v1_final.csv.", "question": "Which file contains the Saudi Skills Taxonomy?", "answers": {"text": ["The taxonomy is in saudi_skills_taxonomy_v1_final.csv."], "answer_start": [0]}}
{"id": "project-010", "title": "Content Tagging and Competency Mapping", "context": "No. They are project, data and evaluation files, not a ready-made QA dataset.", "question": "Are the project data files a ready-made QA dataset?", "answers": {"text": ["No. They are project, data and evaluation files, not a ready-made QA dataset."], "answer_start": [0]}}
{"id": "project-011", "title": "Content Tagging and Competency Mapping", "context": "Qwen2.5-7B-Instruct is used for topic tagging, semantic summarization, and final competency selection and validation.", "question": "Which model is used for topic tagging, semantic summarization and final competency selection?", "answers": {"text": ["Qwen2.5-7B-Instruct is used for topic tagging, semantic summarization, and final competency selection and validation."], "answer_start": [0]}}
{"id": "project-012", "title": "Content Tagging and Competency Mapping", "context": "It uses 4-bit NF4 quantization.", "question": "What quantization does Qwen2.5-7B-Instruct use in this project?", "answers": {"text": ["It uses 4-bit NF4 quantization."], "answer_start": [0]}}
{"id": "project-013", "title": "Content Tagging and Competency Mapping", "context": "intfloat/multilingual-e5-base (E5) performs semantic competency retrieval.", "question": "Which model performs semantic competency retrieval?", "answers": {"text": ["intfloat/multilingual-e5-base (E5) performs semantic competency retrieval."], "answer_start": [0]}}
{"id": "project-014", "title": "Content Tagging and Competency Mapping", "context": "Content is divided into 6,000-character chunks.", "question": "What is the chunk size used to divide content?", "answers": {"text": ["Content is divided into 6,000-character chunks."], "answer_start": [0]}}
{"id": "project-015", "title": "Content Tagging and Competency Mapping", "context": "Chunk overlap is 500 characters.", "question": "How much do chunks overlap?", "answers": {"text": ["Chunk overlap is 500 characters."], "answer_start": [0]}}
{"id": "project-016", "title": "Content Tagging and Competency Mapping", "context": "Lexical matching is also used alongside E5 semantic retrieval.", "question": "Besides E5 semantic retrieval, what other matching is used in retrieval?", "answers": {"text": ["Lexical matching is also used alongside E5 semantic retrieval."], "answer_start": [0]}}
{"id": "project-017", "title": "Content Tagging and Competency Mapping", "context": "The E5 semantic score weight is 0.6 and the lexical score weight is 0.4.", "question": "What are the weights of the semantic and lexical scores?", "answers": {"text": ["The E5 semantic score weight is 0.6 and the lexical score weight is 0.4."], "answer_start": [0]}}
{"id": "project-018", "title": "Content Tagging and Competency Mapping", "context": "The Top-5 candidates are retrieved per chunk.", "question": "How many candidates are retrieved per chunk?", "answers": {"text": ["The Top-5 candidates are retrieved per chunk."], "answer_start": [0]}}
{"id": "project-019", "title": "Content Tagging and Competency Mapping", "context": "Candidates are aggregated across chunks, and a final Top-10 candidate pool is formed.", "question": "How is the final candidate pool formed?", "answers": {"text": ["Candidates are aggregated across chunks, and a final Top-10 candidate pool is formed."], "answer_start": [0]}}
{"id": "project-020", "title": "Content Tagging and Competency Mapping", "context": "Qwen selects and validates the final competencies from the Top-10 pool.", "question": "What does Qwen do after the Top-10 candidate pool is formed?", "answers": {"text": ["Qwen selects and validates the final competencies from the Top-10 pool."], "answer_start": [0]}}
{"id": "project-021", "title": "Content Tagging and Competency Mapping", "context": "At Top-3: Micro Precision 63.89%, Micro Recall 67.65%, and Micro F1 65.71%.", "question": "What are the final competency-mapping Micro Precision, Micro Recall and Micro F1 at Top-3?", "answers": {"text": ["At Top-3: Micro Precision 63.89%, Micro Recall 67.65%, and Micro F1 65.71%."], "answer_start": [0]}}
{"id": "project-022", "title": "Content Tagging and Competency Mapping", "context": "At Top-1: Micro Precision 83.33%, Micro Recall 29.41%, and Micro F1 43.48%.", "question": "What are the final competency-mapping results at Top-1 for micro precision, micro recall and micro F1?", "answers": {"text": ["At Top-1: Micro Precision 83.33%, Micro Recall 29.41%, and Micro F1 43.48%."], "answer_start": [0]}}
{"id": "project-023", "title": "Content Tagging and Competency Mapping", "context": "At Top-5: Micro Precision 45.00%, Micro Recall 79.41%, and Micro F1 57.45%.", "question": "What are the final competency-mapping results at Top-5 for micro precision, micro recall and micro F1?", "answers": {"text": ["At Top-5: Micro Precision 45.00%, Micro Recall 79.41%, and Micro F1 57.45%."], "answer_start": [0]}}
{"id": "project-024", "title": "Content Tagging and Competency Mapping", "context": "There were 23 correct matches at Top-3 and 27 correct matches at Top-5.", "question": "How many correct matches were found at Top-3 and at Top-5?", "answers": {"text": ["There were 23 correct matches at Top-3 and 27 correct matches at Top-5."], "answer_start": [0]}}
{"id": "project-025", "title": "Content Tagging and Competency Mapping", "context": "Macro Recall is 33.33% at Top-1, 70.14% at Top-3, and 81.25% at Top-5.", "question": "What is the Macro Recall at Top-1, Top-3 and Top-5?", "answers": {"text": ["Macro Recall is 33.33% at Top-1, 70.14% at Top-3, and 81.25% at Top-5."], "answer_start": [0]}}
{"id": "project-026", "title": "Content Tagging and Competency Mapping", "context": "Top-3 has the highest Micro F1, at 65.71%.", "question": "At which K does the final mapping reach the highest Micro F1?", "answers": {"text": ["Top-3 has the highest Micro F1, at 65.71%."], "answer_start": [0]}}
{"id": "project-027", "title": "Content Tagging and Competency Mapping", "context": "Hit@1 is 83.33%, Hit@3 is 100.00%, and Hit@5 is 100.00%.", "question": "What are the retrieval Hit@1, Hit@3 and Hit@5 values?", "answers": {"text": ["Hit@1 is 83.33%, Hit@3 is 100.00%, and Hit@5 is 100.00%."], "answer_start": [0]}}
{"id": "project-028", "title": "Content Tagging and Competency Mapping", "context": "Precision@5 is 45.00%, Recall@5 is 81.25%, and F1@5 is 56.85%.", "question": "What are the retrieval Precision@5, Recall@5 and F1@5?", "answers": {"text": ["Precision@5 is 45.00%, Recall@5 is 81.25%, and F1@5 is 56.85%."], "answer_start": [0]}}
{"id": "project-029", "title": "Content Tagging and Competency Mapping", "context": "The MRR is 90.28%.", "question": "What is the retrieval MRR?", "answers": {"text": ["The MRR is 90.28%."], "answer_start": [0]}}
{"id": "project-030", "title": "Content Tagging and Competency Mapping", "context": "No. Hit@3 is a retrieval metric. The final competency-mapping metrics are the Top-K micro precision, micro recall, micro F1, macro recall and correct matches.", "question": "Is the Hit@3 value of 100.00% a final competency-mapping metric?", "answers": {"text": ["No. Hit@3 is a retrieval metric. The final competency-mapping metrics are the Top-K micro precision, micro recall, micro F1, macro recall and correct matches."], "answer_start": [0]}}
{"id": "project-031", "title": "Content Tagging and Competency Mapping", "context": "Micro precision decreases from 83.33% at Top-1 to 63.89% at Top-3 and 45.00% at Top-5, while micro recall increases from 29.41% to 67.65% to 79.41%. Returning more candidates finds more correct competencies but includes more incorrect ones.", "question": "How does precision change as K increases in the final mapping results, and what does that suggest?", "answers": {"text": ["Micro precision decreases from 83.33% at Top-1 to 63.89% at Top-3 and 45.00% at Top-5, while micro recall increases from 29.41% to 67.65% to 79.41%. Returning more candidates finds more correct competencies but includes more incorrect ones."], "answer_start": [0]}}
{"id": "project-032", "title": "Content Tagging and Competency Mapping", "context": "Top-3 has the highest Micro F1 (65.71%), balancing Micro Precision (63.89%) and Micro Recall (67.65%).", "question": "Why is Top-3 considered the best balance in the final mapping results?", "answers": {"text": ["Top-3 has the highest Micro F1 (65.71%), balancing Micro Precision (63.89%) and Micro Recall (67.65%)."], "answer_start": [0]}}
{"id": "project-033", "title": "Content Tagging and Competency Mapping", "context": "It is containerized and deployed using Docker and Kubernetes with GPU model serving, and connected to a working AI Hub interface.", "question": "How is the project deployed?", "answers": {"text": ["It is containerized and deployed using Docker and Kubernetes with GPU model serving, and connected to a working AI Hub interface."], "answer_start": [0]}}
{"id": "project-034", "title": "Content Tagging and Competency Mapping", "context": "User → AI Hub Interface → API / Model Serving → AI Pipeline → Structured Results.", "question": "What is the high-level flow of the deployed system?", "answers": {"text": ["User → AI Hub Interface → API / Model Serving → AI Pipeline → Structured Results."], "answer_start": [0]}}
{"id": "project-035", "title": "Content Tagging and Competency Mapping", "context": "The evaluation set is small (12 files, 34 validated mappings), Top-1 recall is low (29.41%), and precision drops as K increases. The project materials do not provide an explicit limitations list.", "question": "What limitations can be observed from the reported evaluation results?", "answers": {"text": ["The evaluation set is small (12 files, 34 validated mappings), Top-1 recall is low (29.41%), and precision drops as K increases. The project materials do not provide an explicit limitations list."], "answer_start": [0]}}
{"id": "project-036", "title": "Content Tagging and Competency Mapping", "context": "This information is not specified in the project materials. Specific planned future work is not specified.", "question": "What future work is planned for the project?", "answers": {"text": ["This information is not specified in the project materials. Specific planned future work is not specified."], "answer_start": [0]}}
{"id": "project-037", "title": "Content Tagging and Competency Mapping", "context": "This information is not specified in the project materials.", "question": "Which GPU type is used for model serving?", "answers": {"text": ["This information is not specified in the project materials."], "answer_start": [0]}}