20260625:很多新内容

concepts/native-streaming-ar-training.md

@@ -0,0 +1,30 @@
+---
+title: "Native Streaming AR Training"
+created: 2026-06-20
+updated: 2026-06-20
+type: concept
+tags: ["training", "autoregressive", "streaming", "causal"]
+sources: ["https://arxiv.org/abs/2606.17800"]
+---
+
+# Native Streaming AR Training (原生流式自回归训练)
+
+**Native Streaming AR Training** 是 [[maineCoon|MaineCoon]] 的核心训练范式：在训练和推理时使用**相同的因果逐块自回归 regime**，而非通过 teacher forcing 从非因果教师蒸馏。
+
+## 核心原则
+
+- **Chunk-by-chunk causal rollout**：每次预测一个 chunk，仅以已生成的 chunk 为条件
+- **无 Teacher Forcing**：不从双向教师蒸馏流式行为——原生即流式
+- **Train-inference matched**：训练和推理分布一致，消除 gap
+
+## 关键组件
+
+- [[self-resampling|Self-Resampling]]：以模型自身退化历史为条件
+- [[flow-matching|Flow Matching]] loss
+- [[audio-visual-representation-alignment|Cross-Modal Representation Alignment]] 加速
+
+## 参考
+- [[maineCoon|MaineCoon 论文]] Section 3.1
+- [[self-resampling|Self-Resampling]]
+- [[autoregressive-video-generation|自回归视频生成]]
+- [[wan-streamer]] — 端到端流式全双工交互中的原生流式训练