Reference Paper

papers/RefPaperByMarkyan04/Topic4 Discrete diffusion and mixed-type generation/Diffusion-LM Improves Controllable Text Generation/Diffusion-LM Improves Controllable Text Generation.bib

@@ -0,0 +1,9 @@
+@misc{li2022diffusionlmimprovescontrollabletext,
+      title={Diffusion-LM Improves Controllable Text Generation}, 
+      author={Xiang Lisa Li and John Thickstun and Ishaan Gulrajani and Percy Liang and Tatsunori B. Hashimoto},
+      year={2022},
+      eprint={2205.14217},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={httpsarxiv.orgabs2205.14217}, 
+}

papers/RefPaperByMarkyan04/Topic4 Discrete diffusion and mixed-type generation/Intro.txt

@@ -0,0 +1,17 @@
+离散/混合变量扩散（解决 Modbus 的功能码/地址等离散字段）
+
+你是“语义级生成”：至少包含
+离散：Function Code、（可选）异常码、寄存器地址/地址簇、读写长度
+连续/整数：寄存器值、时间间隔（inter-arrival）
+这类“混合类型生成”往往要引用离散扩散或 embedding trick。
+
+Austin et al. Structured Denoising Diffusion Models in Discrete State-Spaces (D3PM). NeurIPS 2021.
+用途：离散状态空间扩散；功能码/异常码/地址簇（token）可以用 D3PM 直接扩散生成。
+
+Li et al. Diffusion-LM Improves Controllable Text Generation. NeurIPS 2022.
+用途：文本是离散序列生成；你可将“PDU 字段序列/行为序列”类比为句子，并用其“可控生成”的讨论支撑“协议约束/场景约束”的必要性。
+
+工程落地常见做法：
+离散字段：D3PM/Multinomial diffusion 或先 embedding 到连续空间再做高斯扩散；
+连续字段：标准 DDPM；
+最后用一个 deterministic assembler（pymodbus/scapy）组装 MBAP/PDU，确保协议有效率接近 100%。

papers/RefPaperByMarkyan04/Topic4 Discrete diffusion and mixed-type generation/Structured Denoising Diffusion Models in Discrete State-Spaces/Structured Denoising Diffusion Models in Discrete State-Spaces.bib

@@ -0,0 +1,9 @@
+@misc{austin2023structureddenoisingdiffusionmodels,
+      title={Structured Denoising Diffusion Models in Discrete State-Spaces}, 
+      author={Jacob Austin and Daniel D. Johnson and Jonathan Ho and Daniel Tarlow and Rianne van den Berg},
+      year={2023},
+      eprint={2107.03006},
+      archivePrefix={arXiv},
+      primaryClass={cs.LG},
+      url={https://arxiv.org/abs/2107.03006}, 
+}