Reading List

Appendix E · Reading List [DRAFT]¶

[To write: Reading List.]

Planned contents¶

• Scheduling & batching: Orca (iteration-level scheduling), Sarathi-Serve (chunked prefill)

• Memory: vLLM / PagedAttention, SGLang / RadixAttention

• Attention: FlashAttention 1–3, Multi-Query Attention, GQA, DeepSeek MLA

• Disaggregation: DistServe, Splitwise

• Speculation: Leviathan et al., Chen et al., Medusa, EAGLE

• Quantisation: LLM.int8(), SmoothQuant, GPTQ, AWQ, FP8 formats

• Multi-tenancy: S-LoRA, Punica

• Structured output: Outlines, XGrammar

• Grouped by the chapter that cites them, so the list doubles as a per-chapter bibliography.