LOCAL PREVIEW View on GitHub

01: FM Output Quality Assessment — Deep Dive Scenarios

Parent: 01-fm-output-quality-assessment.md AIP-C01 Skill: 5.1.1 — Assess the quality of FM outputs (for example, relevance, coherence, fluency, and factual accuracy)


Scenarios

# Scenario Focus Area Difficulty Range
01 Recommendation Relevance Scoring Relevance metrics for manga recommendations Easy → Very Hard
02 FAQ Factual Accuracy Validation Factual accuracy checking for product/policy info Easy → Very Hard
03 Multi-Turn Consistency Check Consistency across conversation turns Easy → Very Hard
04 Japanese Content Fluency Evaluation Fluency evaluation for bilingual content Easy → Very Hard

How to Use These Scenarios

Each scenario folder contains: - README.md — Follow-up questions at 4 difficulty levels (Easy, Medium, Hard, Very Hard) - ANSWERS.md — Detailed answer key with reasoning, code snippets, and MangaAssist context

Interview Prep Flow: 1. Read the parent file for the conceptual foundation 2. Work through Easy questions to confirm understanding 3. Progress to Medium/Hard for system design depth 4. Tackle Very Hard for production edge cases and tradeoffs