agentic-rl-analysis/mkdocs.yml at main · XavierZhang2002/agentic-rl-analysis · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
site_name: "LLM Post-Training 与 Agentic RL 研究"
site_description: "系统梳理大语言模型后训练与 Agentic RL 的技术报告"
site_author: "Xavier"
site_url: https://XavierZhang2002.github.io/agentic-rl-analysis

repo_name: "agentic-rl-analysis"
repo_url: https://github.com/XavierZhang2002/agentic-rl-analysis
edit_uri: edit/main/docs/

theme:
  name: material
  language: zh
  palette:
    - media: "(prefers-color-scheme: light)"
      scheme: default
      primary: indigo
      accent: indigo
      toggle:
        icon: material/brightness-7
        name: 切换到暗色模式
    - media: "(prefers-color-scheme: dark)"
      scheme: slate
      primary: indigo
      accent: indigo
      toggle:
        icon: material/brightness-4
        name: 切换到亮色模式
  font:
    text: Noto Sans SC
    code: JetBrains Mono
  features:
    - navigation.tabs
    - navigation.sections
    - navigation.expand
    - navigation.top
    - navigation.footer
    - navigation.path
    - navigation.instant
    - navigation.instant.progress
    - search.suggest
    - search.highlight
    - content.code.copy
    - content.code.annotate

extra:
  social:
    - icon: fontawesome/brands/github
      link: https://github.com/XavierZhang2002

nav:
  - 首页: index.md
  - Post-Training 技术报告:
      - 概述: post-training/index.md
      - 1. 算法基础:
          - 1.1 大模型训练全景: post-training/ch1/1.1-training-landscape.md
          - 1.2 RLHF 与 RLVR 两大范式: post-training/ch1/1.2-rlhf-rlvr.md
          - 1.3 DPO — 离线偏好优化: post-training/ch1/1.3-dpo.md
          - 1.4 PPO — 经典策略优化: post-training/ch1/1.4-ppo.md
          - 1.5 GRPO — RLVR 奠基算法: post-training/ch1/1.5-grpo.md
          - 1.6 DAPO — 工业级优化: post-training/ch1/1.6-dapo.md
          - 1.7 VAPO — Value 路线回归: post-training/ch1/1.7-vapo.md
          - 1.8 CISPO — 序列级裁剪: post-training/ch1/1.8-cispo.md
          - 1.9 GSPO — 序列级策略优化: post-training/ch1/1.9-gspo.md
          - 1.10 SAPO — 平滑门控: post-training/ch1/1.10-sapo.md
          - 1.11 算法公式速查表: post-training/ch1/1.11-cheatsheet.md
          - 1.12 演进逻辑总结: post-training/ch1/1.12-evolution.md
      - 2. 技术报告深度解读:
          - 2.1 DeepSeek: post-training/ch2/2.1-deepseek.md
          - 2.2 Kimi: post-training/ch2/2.2-kimi.md
          - 2.3 Qwen: post-training/ch2/2.3-qwen.md
          - 2.4 MiniMax: post-training/ch2/2.4-minimax.md
          - 2.5 GLM-5: post-training/ch2/2.5-glm.md
          - 2.6 Seed: post-training/ch2/2.6-seed.md
          - 2.7 闭源模型概览: post-training/ch2/2.7-closed-source.md
          - 2.8 跨模型训练经验: post-training/ch2/2.8-cross-model.md
          - 2.9 数据工程专题: post-training/ch2/2.9-data-engineering.md
          - 2.10 基座的 Agentic 能力是怎么训出来的: post-training/ch2/2.10-agentic-training.md
      - 3. 演进路线与趋势展望:
          - 3.1 技术演进与范式变迁: post-training/ch3/3.1-timeline-paradigms.md
          - 3.2 行业共识与核心挑战: post-training/ch3/3.2-challenges-future.md
          - 3.3 个人分析与思考: post-training/ch3/3.3-opinions.md
  - Agentic RL 调研报告:
      - 概述: agentic-rl/index.md
      - 1. 算法与核心挑战:
          - 1.1 从推理 RL 到 Agentic RL: agentic-rl/ch1/1.1-overview.md
          - 1.2 奖励信号与训练稳定性: agentic-rl/ch1/1.2-reward-stability.md
          - 1.3 探索效率与信用分配: agentic-rl/ch1/1.3-exploration-credit.md
          - 1.4 工程实践: agentic-rl/ch1/1.4-engineering.md
          - 1.5 算法速查与路线图: agentic-rl/ch1/1.5-algorithm-summary.md
      - 2. 全景分析与展望:
          - 2.1 领域全景与技术路线: agentic-rl/ch2/2.1-landscape.md
          - 2.2 核心判断与未来预测: agentic-rl/ch2/2.2-outlook.md

markdown_extensions:
  - admonition
  - pymdownx.details
  - pymdownx.superfences:
      custom_fences:
        - name: mermaid
          class: mermaid-custom
          format: !!python/name:pymdownx.superfences.fence_code_format
  - pymdownx.highlight:
      anchor_linenums: true
  - pymdownx.inlinehilite
  - pymdownx.snippets
  - pymdownx.tabbed:
      alternate_style: true
  - pymdownx.arithmatex:
      generic: true
  - toc:
      permalink: true
      slugify: !!python/name:pymdownx.slugs.uslugify
  - attr_list
  - md_in_html
  - tables
  - footnotes
  - pymdownx.tasklist:
      custom_checkbox: true
  - pymdownx.emoji:
      emoji_index: !!python/name:material.extensions.emoji.twemoji
      emoji_generator: !!python/name:material.extensions.emoji.to_svg

extra_javascript:
  - javascripts/mermaid.mjs
  - javascripts/mathjax.js
  - https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js

extra_css:
  - stylesheets/extra.css

plugins:
  - search
  - tags