A municipal bus-stop snow-clear pack used last year’s cases that almost never included curb-cut routes; after fine-tuning, new packs still skip those streets. Which fine-tuning challenge does this show?
Select an answer to reveal the explanation.
Short Explanation
If the teaching set rarely shows curb-cuts, the tuned model will keep skipping them. High-quality, task-specific data matters; thin or biased sets produce thin or biased results. That is a fine-tuning data challenge, not the Chapter 3 output-bias catalog by itself.
Full Explanation
Fine-tuning depends on high-quality, task-specific training data. A municipal bus-stop snow-clear pack tuned on last year’s cases that almost never included curb-cut routes will keep skipping those streets because a biased or incomplete set yields biased or inaccurate tuned results—the data-quality and bias challenge for fine-tuning in civic winter operations. Accessibility and curb-cut coverage must be present in the labeled examples if the tuned model is expected to generate them. Context-window chunking size selection for RAG handbooks is a retrieval preparation concern, not a tuning-set composition issue. Front-end color theming of the tester form is unrelated UI polish that does not fix missing curb-cut stories. Vendor SKU licensing for a commercial chatbot pane is a commercial or selection topic, not the fine-tuning data challenge shown. Exam caveat: Chapter 3 covers bias as an LLM output defect class; this item focuses on bias and thinness in the tuning data that produce those outputs after fine-tuning. Operational check: before fine-tuning snow-clear packs, audit whether curb-cut and accessibility routes are adequately represented in the labeled set; expand and rebalance the set if those streets are missing.