{"schema_version":"1.0","language":"zh-CN","generated_from":"NEWVAR 从零开始学AI structured Markdown","write_access":false,"full_text_access":false,"content_boundary":"This endpoint exposes metadata, objectives, prerequisites, labs and source summaries. It does not expose chapter body text or answer text.","chapter":{"slug":"flows-energy-evaluation","title":"流模型、能量模型与生成评估","english_title":"Flows, energy models and evaluation","chapter_number":64,"volume":3,"volume_title":"第三卷：主要AI分支","month":16,"unit_title":"生成模型、扩散与流模型","review_status":"draft","updated_at":"2026-08-06","estimated_hours":8,"focus":"理解可逆变换、连续流、能量函数和质量/多样性/记忆风险。","objectives":["理解可逆变换、连续流、能量函数和质量/多样性/记忆风险。","解释核心数学关系并完成最小实现","运行单变量实验并记录失败样例","完成工程线或研究线至少一项迁移任务"],"prerequisites":["diffusion-score"],"exercise_count":3,"tracks":["engineering","research"],"human_url":"/learn/flows-energy-evaluation"},"labs":[{"id":"lab-flows-energy-evaluation","chapter_slug":"flows-energy-evaluation","title":"流模型、能量模型与生成评估：单变量实验","kind":"guidance_scale","data_status":"教学模拟","variable":{"label":"引导强度","unit":"","min":0,"max":20,"step":0.5,"default":7.5},"objective":"只改变“引导强度”，观察结果、代价和风险如何一起变化，理解理解可逆变换、连续流、能量函数和质量/多样性/记忆风险。","static_fallback":"静态替代：把引导强度分别设为0、7.5和20，手工比较三组教学模拟输出。","keyboard":"聚焦滑块后使用方向键微调，Page Up/Page Down大步调整，Home/End到达边界。","measurement_note":"页面数值由公开公式生成，只用于教学，不代表真实模型性能或现实世界因果效果。","source_ids":["S035","S036","S037","S038","S039"]}],"project":{"id":"project-16","month":16,"title":"生成并审计一组教学模拟图案","brief":"比较VAE、GAN、扩散或流的玩具实现，记录模式覆盖、质量、成本、许可和滥用风险。","deliverables":["问题与边界说明","可运行最小实现","实验记录与失败分析","来源与许可清单","风险、隐私与人工闸门","复现README"]},"sources":[{"id":"S035","author":"Diederik P. Kingma, Max Welling","title":"Auto-Encoding Variational Bayes","source_level":"同行评审论文预印本","published_at":"2013","accessed_at":"2026-08-06","doi_or_url":"https://arxiv.org/abs/1312.6114"},{"id":"S036","author":"Ian Goodfellow et al.","title":"Generative Adversarial Nets","source_level":"同行评审论文预印本","published_at":"2014","accessed_at":"2026-08-06","doi_or_url":"https://arxiv.org/abs/1406.2661"},{"id":"S037","author":"Jonathan Ho et al.","title":"Denoising Diffusion Probabilistic Models","source_level":"同行评审论文预印本","published_at":"2020","accessed_at":"2026-08-06","doi_or_url":"https://arxiv.org/abs/2006.11239"},{"id":"S038","author":"Yang Song et al.","title":"Score-Based Generative Modeling through Stochastic Differential Equations","source_level":"同行评审论文预印本","published_at":"2020","accessed_at":"2026-08-06","doi_or_url":"https://arxiv.org/abs/2011.13456"},{"id":"S039","author":"Yaron Lipman et al.","title":"Flow Matching for Generative Modeling","source_level":"同行评审论文预印本","published_at":"2022","accessed_at":"2026-08-06","doi_or_url":"https://arxiv.org/abs/2210.02747"}]}