-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathgraph.json
More file actions
260 lines (241 loc) · 29.4 KB
/
Copy pathgraph.json
File metadata and controls
260 lines (241 loc) · 29.4 KB
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
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
{
"$comment": "Single source of truth for the atlas. Every node MUST have a corresponding card under docs/data/cards/ (concept nodes share concepts.md). Validator: tools/validate_graph.py",
"nodes": [
{"id": "paper:2212.10156", "label": "UniAD", "label_zh": "UniAD（统一规划导向 AD）", "kind": "paper", "tier": "spine", "topic": "e2e_ad", "phase": "core", "year": 2022, "card": "paper_2212.10156_uniad.md", "labs": ["lab:lab03"]},
{"id": "paper:2210.14222", "label": "PlanT", "label_zh": "PlanT（对象级规划 transformer）", "kind": "paper", "tier": "spine", "topic": "e2e_ad", "phase": "core", "year": 2022, "card": "paper_2210.14222_plant.md", "labs": ["lab:lab04"]},
{"id": "paper:2402.12289", "label": "DriveVLM", "label_zh": "DriveVLM / DriveVLM-Dual", "kind": "paper", "tier": "spine", "topic": "vlm_vla", "phase": "frontier", "year": 2024, "card": "paper_2402.12289_drivevlm.md", "labs": ["lab:lab09"]},
{"id": "paper:2311.10813", "label": "Agent-Driver", "label_zh": "Agent-Driver（LLM 认知 agent）", "kind": "paper", "tier": "spine", "topic": "vlm_vla", "phase": "frontier", "year": 2023, "card": "paper_2311.10813_agent_driver.md", "labs": ["lab:lab08"]},
{"id": "paper:2309.16292", "label": "DiLu", "label_zh": "DiLu（知识驱动 + 反思）", "kind": "paper", "tier": "spine", "topic": "vlm_vla", "phase": "frontier", "year": 2023, "card": "paper_2309.16292_dilu.md", "labs": ["lab:lab07"]},
{"id": "paper:2307.01694", "label": "Spike-driven Transformer", "label_zh": "Spike-driven Transformer", "kind": "paper", "tier": "spine", "topic": "brain_inspired", "phase": "frontier", "year": 2023, "card": "paper_2307.01694_spike_driven_transformer.md", "labs": ["lab:lab06"]},
{"id": "paper:2508.10104", "label": "DINOv3", "label_zh": "DINOv3", "kind": "paper", "tier": "spine", "topic": "ssl_vision", "phase": "frontier", "year": 2025, "card": "paper_2508.10104_dinov3.md", "labs": ["lab:lab05"]},
{"id": "paper:2512.24426", "label": "CF-VLA", "label_zh": "CF-VLA（反事实 VLA）", "kind": "paper", "tier": "spine", "topic": "vlm_vla", "phase": "frontier", "year": 2025, "card": "paper_2512.24426_cfvla.md", "labs": ["lab:lab10"]},
{"id": "channel:3blue1brown", "label": "3Blue1Brown", "label_zh": "3Blue1Brown（可视化数学）", "kind": "channel", "tier": "spine", "topic": "math_foundations", "phase": "prereq", "year": 2015, "card": "channel_3blue1brown.md"},
{"id": "channel:mu_li_bilibili", "label": "跟李沐学AI", "label_zh": "跟李沐学AI（B 站论文精读）", "kind": "channel", "tier": "spine", "topic": "companion_media", "phase": "prereq", "year": 2020, "card": "channel_mu_li_bilibili.md"},
{"id": "channel:ez_encoder_academy", "label": "EZ.Encoder", "label_zh": "EZ.Encoder Academy", "kind": "channel", "tier": "spine", "topic": "companion_media", "phase": "prereq", "year": 2023, "card": "channel_ez_encoder_academy.md"},
{"id": "course:zhao_rl", "label": "Zhao Shiyu RL", "label_zh": "赵世钰 强化学习的数学原理", "kind": "course", "tier": "spine", "topic": "rl_foundations", "phase": "prereq", "year": 2024, "card": "course_zhao_shiyu_rl.md", "labs": ["lab:lab01"]},
{"id": "course:cs285", "label": "CS285 Deep RL", "label_zh": "Berkeley CS285 深度强化学习", "kind": "course", "tier": "spine", "topic": "deep_rl", "phase": "core", "year": 2023, "card": "course_cs285_levine.md", "labs": ["lab:lab02"]},
{"id": "essay:bitter_lesson", "label": "Bitter Lesson", "label_zh": "苦涩的教训（Sutton）", "kind": "essay", "tier": "spine", "topic": "meta_philosophy", "phase": "prereq", "year": 2019, "card": "essay_bitter_lesson.md"},
{"id": "paper:vaswani2017", "label": "Transformer", "label_zh": "Attention Is All You Need", "kind": "paper", "tier": "S", "topic": "math_foundations", "phase": "prereq", "year": 2017, "card": "paper_vaswani2017.md"},
{"id": "paper:vit", "label": "ViT", "label_zh": "ViT（图像 transformer）", "kind": "paper", "tier": "S", "topic": "ssl_vision", "phase": "prereq", "year": 2020, "card": "paper_vit.md"},
{"id": "paper:carion2020", "label": "DETR", "label_zh": "DETR（端到端检测）", "kind": "paper", "tier": "S", "topic": "ssl_vision", "phase": "prereq", "year": 2020, "card": "paper_carion2020.md"},
{"id": "paper:he2015_resnet", "label": "ResNet", "label_zh": "ResNet（残差网络）", "kind": "paper", "tier": "S", "topic": "ssl_vision", "phase": "prereq", "year": 2015, "card": "paper_he2015_resnet.md"},
{"id": "paper:gpt3", "label": "GPT-3", "label_zh": "GPT-3（few-shot 大语言模型）", "kind": "paper", "tier": "S", "topic": "vlm_vla", "phase": "prereq", "year": 2020, "card": "paper_gpt3.md"},
{"id": "paper:schulman2017_ppo", "label": "PPO", "label_zh": "PPO（近端策略优化）", "kind": "paper", "tier": "S", "topic": "deep_rl", "phase": "core", "year": 2017, "card": "paper_schulman2017_ppo.md"},
{"id": "paper:mnih2015_dqn", "label": "DQN", "label_zh": "DQN（深度 Q 网络）", "kind": "paper", "tier": "S", "topic": "deep_rl", "phase": "prereq", "year": 2015, "card": "paper_mnih2015_dqn.md"},
{"id": "paper:ross2011_dagger", "label": "DAgger", "label_zh": "DAgger（解决协变量偏移）", "kind": "paper", "tier": "S", "topic": "deep_rl", "phase": "core", "year": 2011, "card": "paper_ross2011_dagger.md", "labs": ["lab:lab02"]},
{"id": "paper:silver2017_alphazero", "label": "AlphaGo Zero", "label_zh": "AlphaGo Zero / AlphaZero", "kind": "paper", "tier": "S", "topic": "deep_rl", "phase": "prereq", "year": 2017, "card": "paper_silver2017_alphazero.md"},
{"id": "paper:li2022bevformer", "label": "BEVFormer", "label_zh": "BEVFormer（时空 BEV transformer）", "kind": "paper", "tier": "S", "topic": "e2e_ad", "phase": "prereq", "year": 2022, "card": "paper_li2022bevformer.md"},
{"id": "paper:vadv2", "label": "VADv2", "label_zh": "VADv2（向量化端到端 + 概率规划）", "kind": "paper", "tier": "A", "topic": "e2e_ad", "phase": "frontier", "year": 2024, "card": "paper_vadv2.md"},
{"id": "paper:transfuser", "label": "TransFuser", "label_zh": "TransFuser（多模态融合 BC）", "kind": "paper", "tier": "A", "topic": "e2e_ad", "phase": "core", "year": 2022, "card": "paper_transfuser.md"},
{"id": "paper:llava", "label": "LLaVA / Qwen-VL", "label_zh": "LLaVA / Qwen-VL（开源 VLM）", "kind": "paper", "tier": "A", "topic": "vlm_vla", "phase": "prereq", "year": 2023, "card": "paper_llava.md"},
{"id": "paper:gaia1", "label": "GAIA-1", "label_zh": "GAIA-1（生成式驾驶世界模型）", "kind": "paper", "tier": "A", "topic": "vlm_vla", "phase": "frontier", "year": 2023, "card": "paper_gaia1.md"},
{"id": "paper:drivedreamer", "label": "DriveDreamer", "label_zh": "DriveDreamer（视频扩散世界模型）", "kind": "paper", "tier": "A", "topic": "vlm_vla", "phase": "frontier", "year": 2024, "card": "paper_drivedreamer.md"},
{"id": "paper:dinov2", "label": "DINOv2", "label_zh": "DINOv2", "kind": "paper", "tier": "A", "topic": "ssl_vision", "phase": "prereq", "year": 2023, "card": "paper_dinov2.md"},
{"id": "paper:sam", "label": "SAM", "label_zh": "SAM / SAM 2（分割万物）", "kind": "paper", "tier": "A", "topic": "ssl_vision", "phase": "prereq", "year": 2023, "card": "paper_sam.md"},
{"id": "paper:sutton_barto", "label": "Sutton & Barto RL", "label_zh": "Sutton & Barto《强化学习导论》", "kind": "paper", "tier": "A", "topic": "rl_foundations", "phase": "prereq", "year": 2018, "card": "paper_sutton_barto.md"},
{"id": "paper:rlhf_dpo", "label": "RLHF / DPO", "label_zh": "RLHF / DPO（人类偏好对齐）", "kind": "paper", "tier": "A", "topic": "deep_rl", "phase": "core", "year": 2023, "card": "paper_rlhf_dpo.md"},
{"id": "paper:world_models", "label": "World Models", "label_zh": "World Models（Ha & Schmidhuber）", "kind": "paper", "tier": "A", "topic": "deep_rl", "phase": "core", "year": 2018, "card": "paper_world_models.md"},
{"id": "paper:mamba", "label": "Mamba", "label_zh": "Mamba（状态空间模型）", "kind": "paper", "tier": "B", "topic": "math_foundations", "phase": "frontier", "year": 2023, "card": "paper_mamba.md"},
{"id": "paper:diffuser", "label": "Diffuser", "label_zh": "Diffuser / Decision Diffuser", "kind": "paper", "tier": "B", "topic": "deep_rl", "phase": "frontier", "year": 2022, "card": "paper_diffuser.md"},
{"id": "paper:ad_benchmarks", "label": "AD benchmarks", "label_zh": "CARLA / nuScenes / NAVSIM / Bench2Drive", "kind": "paper", "tier": "B", "topic": "e2e_ad", "phase": "prereq", "year": 2017, "card": "benchmarks_ad.md"},
{"id": "paper:lingo2", "label": "Wayve LINGO-2", "label_zh": "Wayve LINGO-2", "kind": "paper", "tier": "B", "topic": "vlm_vla", "phase": "frontier", "year": 2024, "card": "paper_lingo2.md"},
{"id": "paper:tesla_ai_day", "label": "Tesla AI Day", "label_zh": "Tesla AI Day", "kind": "paper", "tier": "B", "topic": "e2e_ad", "phase": "prereq", "year": 2022, "card": "paper_tesla_ai_day.md"},
{"id": "concept:mdp", "label": "MDP", "label_zh": "马尔可夫决策过程 MDP", "kind": "concept", "tier": "concept", "topic": "rl_foundations", "phase": "prereq", "year": 1957, "card": "../../../concepts.md#mdp"},
{"id": "concept:bellman_eq", "label": "Bellman 方程", "label_zh": "Bellman 方程 / 最优方程", "kind": "concept", "tier": "concept", "topic": "rl_foundations", "phase": "prereq", "year": 1957, "card": "../../../concepts.md#bellman方程"},
{"id": "concept:value_iteration", "label": "Value Iteration", "label_zh": "值迭代 / 策略迭代", "kind": "concept", "tier": "concept", "topic": "rl_foundations", "phase": "prereq", "year": 1957, "card": "../../../concepts.md#值迭代--策略迭代"},
{"id": "concept:td_learning", "label": "TD Learning", "label_zh": "TD 学习 / Q-learning / SARSA", "kind": "concept", "tier": "concept", "topic": "rl_foundations", "phase": "prereq", "year": 1988, "card": "../../../concepts.md#td学习--q-learning--sarsa"},
{"id": "concept:policy_gradient", "label": "Policy Gradient", "label_zh": "策略梯度 / REINFORCE", "kind": "concept", "tier": "concept", "topic": "deep_rl", "phase": "core", "year": 1992, "card": "../../../concepts.md#policy-gradient--reinforce"},
{"id": "concept:actor_critic", "label": "Actor-Critic", "label_zh": "Actor-Critic", "kind": "concept", "tier": "concept", "topic": "deep_rl", "phase": "core", "year": 2000, "card": "../../../concepts.md#actor-critic"},
{"id": "concept:ppo", "label": "PPO 概念", "label_zh": "PPO 替代目标", "kind": "concept", "tier": "concept", "topic": "deep_rl", "phase": "core", "year": 2017, "card": "../../../concepts.md#ppo"},
{"id": "concept:dqn", "label": "DQN 概念", "label_zh": "DQN / replay buffer / target net", "kind": "concept", "tier": "concept", "topic": "deep_rl", "phase": "core", "year": 2015, "card": "../../../concepts.md#dqn"},
{"id": "concept:replay_buffer", "label": "Replay buffer", "label_zh": "经验回放", "kind": "concept", "tier": "concept", "topic": "deep_rl", "phase": "core", "year": 1992, "card": "../../../concepts.md#replay-buffer"},
{"id": "concept:imitation_learning", "label": "Imitation Learning", "label_zh": "模仿学习 / BC", "kind": "concept", "tier": "concept", "topic": "deep_rl", "phase": "core", "year": 1989, "card": "../../../concepts.md#模仿学习--bc"},
{"id": "concept:covariate_shift", "label": "Covariate shift", "label_zh": "协变量偏移", "kind": "concept", "tier": "concept", "topic": "deep_rl", "phase": "core", "year": 2010, "card": "../../../concepts.md#协变量偏移"},
{"id": "concept:transformer", "label": "Transformer", "label_zh": "Transformer 架构", "kind": "concept", "tier": "concept", "topic": "math_foundations", "phase": "prereq", "year": 2017, "card": "../../../concepts.md#transformer"},
{"id": "concept:self_attention", "label": "Self-Attention", "label_zh": "Self-Attention", "kind": "concept", "tier": "concept", "topic": "math_foundations", "phase": "prereq", "year": 2017, "card": "../../../concepts.md#self-attention"},
{"id": "concept:detr_query", "label": "DETR query", "label_zh": "DETR object query", "kind": "concept", "tier": "concept", "topic": "ssl_vision", "phase": "prereq", "year": 2020, "card": "../../../concepts.md#detr-query"},
{"id": "concept:bev", "label": "BEV", "label_zh": "BEV 感知", "kind": "concept", "tier": "concept", "topic": "e2e_ad", "phase": "prereq", "year": 2020, "card": "../../../concepts.md#bev感知"},
{"id": "concept:vlm", "label": "VLM", "label_zh": "Vision-Language Model", "kind": "concept", "tier": "concept", "topic": "vlm_vla", "phase": "core", "year": 2021, "card": "../../../concepts.md#vlm"},
{"id": "concept:vla", "label": "VLA", "label_zh": "Vision-Language-Action", "kind": "concept", "tier": "concept", "topic": "vlm_vla", "phase": "frontier", "year": 2023, "card": "../../../concepts.md#vla"},
{"id": "concept:cot", "label": "Chain-of-Thought", "label_zh": "Chain-of-Thought 推理", "kind": "concept", "tier": "concept", "topic": "vlm_vla", "phase": "core", "year": 2022, "card": "../../../concepts.md#chain-of-thought"},
{"id": "concept:tool_use", "label": "Tool use", "label_zh": "Tool use / function calling", "kind": "concept", "tier": "concept", "topic": "vlm_vla", "phase": "core", "year": 2023, "card": "../../../concepts.md#tool-use"},
{"id": "concept:counterfactual", "label": "Counterfactual reasoning", "label_zh": "反事实推理", "kind": "concept", "tier": "concept", "topic": "vlm_vla", "phase": "frontier", "year": 2025, "card": "../../../concepts.md#反事实推理"},
{"id": "concept:meta_action", "label": "Meta-action", "label_zh": "Meta-action / 高层语义动作", "kind": "concept", "tier": "concept", "topic": "vlm_vla", "phase": "frontier", "year": 2024, "card": "../../../concepts.md#meta-action"},
{"id": "concept:ssl", "label": "SSL", "label_zh": "自监督学习", "kind": "concept", "tier": "concept", "topic": "ssl_vision", "phase": "core", "year": 2018, "card": "../../../concepts.md#ssl"},
{"id": "concept:spiking_nn", "label": "Spiking NN", "label_zh": "脉冲神经网络", "kind": "concept", "tier": "concept", "topic": "brain_inspired", "phase": "frontier", "year": 1997, "card": "../../../concepts.md#脉冲神经网络"},
{"id": "concept:rlhf", "label": "RLHF", "label_zh": "RLHF / DPO 概念", "kind": "concept", "tier": "concept", "topic": "deep_rl", "phase": "core", "year": 2017, "card": "../../../concepts.md#rlhf"},
{"id": "concept:scaling_vs_knowledge", "label": "Scaling vs Knowledge", "label_zh": "Scaling vs 人工知识", "kind": "concept", "tier": "concept", "topic": "meta_philosophy", "phase": "prereq", "year": 2019, "card": "../../../concepts.md#scaling-vs-人工知识"},
{"id": "lab:lab00", "label": "lab00 sanity", "label_zh": "lab00 环境检查", "kind": "lab", "tier": "lab", "topic": "math_foundations", "phase": "prereq", "year": 2026, "card": "../../../labs/lab00_environment_check.ipynb"},
{"id": "lab:lab01", "label": "lab01 value iter", "label_zh": "lab01 值迭代 gridworld", "kind": "lab", "tier": "lab", "topic": "rl_foundations", "phase": "prereq", "year": 2026, "card": "../../../labs/lab01_zhao_value_iteration_gridworld.ipynb"},
{"id": "lab:lab02", "label": "lab02 BC vs DAgger", "label_zh": "lab02 BC vs DAgger", "kind": "lab", "tier": "lab", "topic": "deep_rl", "phase": "core", "year": 2026, "card": "../../../labs/lab02_cs285_bc_vs_dagger_minicar.ipynb"},
{"id": "lab:lab03", "label": "lab03 UniAD query", "label_zh": "lab03 UniAD query intuition", "kind": "lab", "tier": "lab", "topic": "e2e_ad", "phase": "core", "year": 2026, "card": "../../../labs/lab03_uniad_query_intuition.ipynb"},
{"id": "lab:lab04", "label": "lab04 PlanT", "label_zh": "lab04 PlanT object planner", "kind": "lab", "tier": "lab", "topic": "e2e_ad", "phase": "core", "year": 2026, "card": "../../../labs/lab04_plant_object_level_planner.ipynb"},
{"id": "lab:lab05", "label": "lab05 DINOv3", "label_zh": "lab05 DINOv3 features", "kind": "lab", "tier": "lab", "topic": "ssl_vision", "phase": "frontier", "year": 2026, "card": "../../../labs/lab05_dinov3_features_minidata.ipynb"},
{"id": "lab:lab06", "label": "lab06 Spike attn", "label_zh": "lab06 Spike-driven attention", "kind": "lab", "tier": "lab", "topic": "brain_inspired", "phase": "frontier", "year": 2026, "card": "../../../labs/lab06_spike_driven_attention_mnist.ipynb"},
{"id": "lab:lab07", "label": "lab07 DiLu loop", "label_zh": "lab07 DiLu LLM 决策循环", "kind": "lab", "tier": "lab", "topic": "vlm_vla", "phase": "frontier", "year": 2026, "card": "../../../labs/lab07_dilu_llm_decision_loop.ipynb"},
{"id": "lab:lab08", "label": "lab08 Agent-Driver tools", "label_zh": "lab08 Agent-Driver tool calling", "kind": "lab", "tier": "lab", "topic": "vlm_vla", "phase": "frontier", "year": 2026, "card": "../../../labs/lab08_agent_driver_tool_calling.ipynb"},
{"id": "lab:lab09", "label": "lab09 DriveVLM-Dual", "label_zh": "lab09 DriveVLM-Dual pipeline", "kind": "lab", "tier": "lab", "topic": "vlm_vla", "phase": "frontier", "year": 2026, "card": "../../../labs/lab09_drivevlm_dual_pipeline.ipynb"},
{"id": "lab:lab10", "label": "lab10 CF-VLA", "label_zh": "lab10 CF-VLA replanner", "kind": "lab", "tier": "lab", "topic": "vlm_vla", "phase": "frontier", "year": 2026, "card": "../../../labs/lab10_cfvla_counterfactual_replanner.ipynb"}
],
"edges": [
{"source": "paper:vaswani2017", "target": "paper:vit", "rel": "prereq"},
{"source": "paper:vaswani2017", "target": "paper:carion2020", "rel": "prereq"},
{"source": "paper:vaswani2017", "target": "paper:gpt3", "rel": "prereq"},
{"source": "paper:vaswani2017", "target": "paper:2307.01694", "rel": "prereq"},
{"source": "paper:vit", "target": "paper:dinov2", "rel": "prereq"},
{"source": "paper:vit", "target": "paper:sam", "rel": "prereq"},
{"source": "paper:vit", "target": "paper:li2022bevformer", "rel": "prereq"},
{"source": "paper:vit", "target": "paper:llava", "rel": "prereq"},
{"source": "paper:vit", "target": "paper:2307.01694", "rel": "prereq"},
{"source": "paper:dinov2", "target": "paper:2508.10104", "rel": "prereq"},
{"source": "paper:carion2020", "target": "paper:li2022bevformer", "rel": "prereq"},
{"source": "paper:carion2020", "target": "paper:2212.10156", "rel": "prereq"},
{"source": "paper:carion2020", "target": "paper:2210.14222", "rel": "prereq"},
{"source": "paper:li2022bevformer", "target": "paper:2212.10156", "rel": "prereq"},
{"source": "paper:he2015_resnet", "target": "paper:vit", "rel": "prereq"},
{"source": "paper:he2015_resnet", "target": "paper:2307.01694", "rel": "prereq"},
{"source": "paper:gpt3", "target": "paper:llava", "rel": "prereq"},
{"source": "paper:gpt3", "target": "paper:rlhf_dpo", "rel": "prereq"},
{"source": "paper:gpt3", "target": "paper:2309.16292", "rel": "prereq"},
{"source": "paper:gpt3", "target": "paper:2311.10813", "rel": "prereq"},
{"source": "paper:llava", "target": "paper:2402.12289", "rel": "prereq"},
{"source": "paper:llava", "target": "paper:2512.24426", "rel": "prereq"},
{"source": "course:zhao_rl", "target": "course:cs285", "rel": "prereq"},
{"source": "course:zhao_rl", "target": "paper:mnih2015_dqn", "rel": "covers"},
{"source": "course:zhao_rl", "target": "paper:schulman2017_ppo", "rel": "covers"},
{"source": "course:zhao_rl", "target": "concept:mdp", "rel": "covers"},
{"source": "course:zhao_rl", "target": "concept:bellman_eq", "rel": "covers"},
{"source": "course:zhao_rl", "target": "concept:value_iteration", "rel": "covers"},
{"source": "course:zhao_rl", "target": "concept:td_learning", "rel": "covers"},
{"source": "course:zhao_rl", "target": "concept:policy_gradient", "rel": "covers"},
{"source": "course:zhao_rl", "target": "concept:actor_critic", "rel": "covers"},
{"source": "course:cs285", "target": "concept:imitation_learning", "rel": "covers"},
{"source": "course:cs285", "target": "concept:covariate_shift", "rel": "covers"},
{"source": "course:cs285", "target": "concept:dqn", "rel": "covers"},
{"source": "course:cs285", "target": "concept:ppo", "rel": "covers"},
{"source": "course:cs285", "target": "concept:replay_buffer", "rel": "covers"},
{"source": "course:cs285", "target": "paper:ross2011_dagger", "rel": "covers"},
{"source": "course:cs285", "target": "paper:rlhf_dpo", "rel": "covers"},
{"source": "course:cs285", "target": "paper:world_models", "rel": "covers"},
{"source": "paper:sutton_barto", "target": "course:zhao_rl", "rel": "parallel"},
{"source": "paper:sutton_barto", "target": "essay:bitter_lesson", "rel": "parallel"},
{"source": "channel:3blue1brown", "target": "concept:transformer", "rel": "covers"},
{"source": "channel:3blue1brown", "target": "concept:self_attention", "rel": "covers"},
{"source": "channel:3blue1brown", "target": "paper:vaswani2017", "rel": "covers"},
{"source": "channel:mu_li_bilibili", "target": "paper:vaswani2017", "rel": "covers"},
{"source": "channel:mu_li_bilibili", "target": "paper:vit", "rel": "covers"},
{"source": "channel:mu_li_bilibili", "target": "paper:carion2020", "rel": "covers"},
{"source": "channel:mu_li_bilibili", "target": "paper:gpt3", "rel": "covers"},
{"source": "channel:mu_li_bilibili", "target": "paper:dinov2", "rel": "covers"},
{"source": "channel:mu_li_bilibili", "target": "paper:he2015_resnet", "rel": "covers"},
{"source": "channel:ez_encoder_academy", "target": "paper:gpt3", "rel": "covers"},
{"source": "channel:ez_encoder_academy", "target": "essay:bitter_lesson", "rel": "covers"},
{"source": "essay:bitter_lesson", "target": "paper:silver2017_alphazero", "rel": "covers"},
{"source": "essay:bitter_lesson", "target": "paper:2309.16292", "rel": "contrasts"},
{"source": "essay:bitter_lesson", "target": "paper:2311.10813", "rel": "contrasts"},
{"source": "essay:bitter_lesson", "target": "paper:2210.14222", "rel": "contrasts"},
{"source": "essay:bitter_lesson", "target": "paper:2307.01694", "rel": "contrasts"},
{"source": "essay:bitter_lesson", "target": "paper:2508.10104", "rel": "covers"},
{"source": "essay:bitter_lesson", "target": "paper:gpt3", "rel": "covers"},
{"source": "paper:2212.10156", "target": "paper:vadv2", "rel": "extends"},
{"source": "paper:2212.10156", "target": "paper:2210.14222", "rel": "contrasts"},
{"source": "paper:2212.10156", "target": "paper:2402.12289", "rel": "feeds"},
{"source": "paper:2210.14222", "target": "paper:transfuser", "rel": "parallel"},
{"source": "paper:2210.14222", "target": "paper:2311.10813", "rel": "feeds"},
{"source": "paper:2402.12289", "target": "paper:2512.24426", "rel": "extends"},
{"source": "paper:2311.10813", "target": "paper:2402.12289", "rel": "extends"},
{"source": "paper:2311.10813", "target": "paper:2309.16292", "rel": "parallel"},
{"source": "paper:2309.16292", "target": "paper:2311.10813", "rel": "parallel"},
{"source": "paper:2508.10104", "target": "paper:2212.10156", "rel": "feeds"},
{"source": "paper:2508.10104", "target": "paper:2402.12289", "rel": "feeds"},
{"source": "paper:2307.01694", "target": "paper:2508.10104", "rel": "contrasts"},
{"source": "paper:gaia1", "target": "paper:2512.24426", "rel": "parallel"},
{"source": "paper:drivedreamer", "target": "paper:gaia1", "rel": "parallel"},
{"source": "paper:world_models", "target": "paper:gaia1", "rel": "prereq"},
{"source": "paper:world_models", "target": "paper:drivedreamer", "rel": "prereq"},
{"source": "paper:world_models", "target": "paper:2512.24426", "rel": "prereq"},
{"source": "paper:rlhf_dpo", "target": "paper:2512.24426", "rel": "prereq"},
{"source": "paper:transfuser", "target": "paper:2210.14222", "rel": "feeds"},
{"source": "paper:vadv2", "target": "paper:2402.12289", "rel": "parallel"},
{"source": "paper:mamba", "target": "paper:vaswani2017", "rel": "contrasts"},
{"source": "paper:diffuser", "target": "paper:vadv2", "rel": "parallel"},
{"source": "paper:lingo2", "target": "paper:2402.12289", "rel": "parallel"},
{"source": "paper:tesla_ai_day", "target": "paper:2212.10156", "rel": "parallel"},
{"source": "paper:ad_benchmarks", "target": "paper:2212.10156", "rel": "covers"},
{"source": "paper:ad_benchmarks", "target": "paper:2402.12289", "rel": "covers"},
{"source": "paper:ad_benchmarks", "target": "paper:2210.14222", "rel": "covers"},
{"source": "paper:ad_benchmarks", "target": "paper:2311.10813", "rel": "covers"},
{"source": "paper:ad_benchmarks", "target": "paper:2309.16292", "rel": "covers"},
{"source": "paper:mnih2015_dqn", "target": "concept:dqn", "rel": "covers"},
{"source": "paper:mnih2015_dqn", "target": "concept:replay_buffer", "rel": "covers"},
{"source": "paper:schulman2017_ppo", "target": "concept:ppo", "rel": "covers"},
{"source": "paper:schulman2017_ppo", "target": "concept:actor_critic", "rel": "covers"},
{"source": "paper:ross2011_dagger", "target": "concept:imitation_learning", "rel": "covers"},
{"source": "paper:ross2011_dagger", "target": "concept:covariate_shift", "rel": "covers"},
{"source": "paper:vaswani2017", "target": "concept:transformer", "rel": "covers"},
{"source": "paper:vaswani2017", "target": "concept:self_attention", "rel": "covers"},
{"source": "paper:carion2020", "target": "concept:detr_query", "rel": "covers"},
{"source": "paper:li2022bevformer", "target": "concept:bev", "rel": "covers"},
{"source": "paper:llava", "target": "concept:vlm", "rel": "covers"},
{"source": "paper:2402.12289", "target": "concept:vla", "rel": "covers"},
{"source": "paper:2402.12289", "target": "concept:cot", "rel": "covers"},
{"source": "paper:2311.10813", "target": "concept:tool_use", "rel": "covers"},
{"source": "paper:2309.16292", "target": "concept:cot", "rel": "covers"},
{"source": "paper:2512.24426", "target": "concept:counterfactual", "rel": "covers"},
{"source": "paper:2512.24426", "target": "concept:meta_action", "rel": "covers"},
{"source": "paper:dinov2", "target": "concept:ssl", "rel": "covers"},
{"source": "paper:2508.10104", "target": "concept:ssl", "rel": "covers"},
{"source": "paper:2307.01694", "target": "concept:spiking_nn", "rel": "covers"},
{"source": "paper:rlhf_dpo", "target": "concept:rlhf", "rel": "covers"},
{"source": "essay:bitter_lesson", "target": "concept:scaling_vs_knowledge", "rel": "covers"},
{"source": "lab:lab01", "target": "course:zhao_rl", "rel": "implements"},
{"source": "lab:lab01", "target": "concept:value_iteration", "rel": "implements"},
{"source": "lab:lab02", "target": "course:cs285", "rel": "implements"},
{"source": "lab:lab02", "target": "paper:ross2011_dagger", "rel": "implements"},
{"source": "lab:lab03", "target": "paper:2212.10156", "rel": "implements"},
{"source": "lab:lab04", "target": "paper:2210.14222", "rel": "implements"},
{"source": "lab:lab05", "target": "paper:2508.10104", "rel": "implements"},
{"source": "lab:lab06", "target": "paper:2307.01694", "rel": "implements"},
{"source": "lab:lab07", "target": "paper:2309.16292", "rel": "implements"},
{"source": "lab:lab08", "target": "paper:2311.10813", "rel": "implements"},
{"source": "lab:lab09", "target": "paper:2402.12289", "rel": "implements"},
{"source": "lab:lab10", "target": "paper:2512.24426", "rel": "implements"}
],
"playbooks": {
"A": {
"title": "Path A — 强化学习从零 / RL from scratch",
"nodes": ["course:zhao_rl", "essay:bitter_lesson", "course:cs285", "paper:ross2011_dagger", "paper:mnih2015_dqn", "paper:schulman2017_ppo", "paper:rlhf_dpo", "paper:2309.16292", "paper:2311.10813", "lab:lab01", "lab:lab02", "lab:lab07", "concept:bellman_eq", "concept:value_iteration", "concept:td_learning", "concept:policy_gradient", "concept:actor_critic", "concept:dqn", "concept:ppo", "concept:imitation_learning", "concept:covariate_shift"]
},
"B": {
"title": "Path B — 感知到端到端 AD / Perception → E2E",
"nodes": ["channel:3blue1brown", "channel:mu_li_bilibili", "paper:vaswani2017", "paper:vit", "paper:he2015_resnet", "paper:carion2020", "paper:li2022bevformer", "paper:dinov2", "paper:2508.10104", "paper:2212.10156", "paper:2210.14222", "paper:vadv2", "paper:transfuser", "paper:2402.12289", "paper:2512.24426", "lab:lab03", "lab:lab04", "lab:lab05", "lab:lab09", "lab:lab10", "concept:transformer", "concept:detr_query", "concept:bev", "concept:vla", "concept:cot"]
},
"C": {
"title": "Path C — 类脑 / 高效 AD / Brain-inspired & efficient",
"nodes": ["paper:vit", "paper:2307.01694", "paper:2508.10104", "paper:2212.10156", "essay:bitter_lesson", "paper:mamba", "lab:lab06", "concept:spiking_nn", "concept:self_attention", "concept:scaling_vs_knowledge"]
},
"D": {
"title": "Path D — LLM·VLM·VLA 范式",
"nodes": ["channel:3blue1brown", "channel:mu_li_bilibili", "paper:vaswani2017", "paper:gpt3", "paper:llava", "paper:rlhf_dpo", "paper:2309.16292", "paper:2311.10813", "paper:2402.12289", "paper:2512.24426", "paper:gaia1", "paper:drivedreamer", "paper:world_models", "paper:lingo2", "lab:lab07", "lab:lab08", "lab:lab09", "lab:lab10", "concept:vlm", "concept:vla", "concept:cot", "concept:tool_use", "concept:counterfactual", "concept:meta_action", "concept:rlhf"]
}
},
"topic_palette": {
"math_foundations": "#7c4dff",
"rl_foundations": "#1565c0",
"deep_rl": "#0d47a1",
"ssl_vision": "#2e7d32",
"e2e_ad": "#ef6c00",
"vlm_vla": "#c62828",
"brain_inspired": "#6a1b9a",
"meta_philosophy": "#424242",
"companion_media": "#00838f"
},
"tier_border_width": {
"spine": 6,
"S": 4,
"A": 2,
"B": 1,
"concept": 1,
"lab": 2
}
}