Pacing Branch Parallelism in LLM Serving
cs.DC, cs.AI, cs.CL
Submitted: 2026-05-07
Updated: 2026-11-01
Code: https://github.com/Azure/AzurePublicDataset
Terminology
Sources
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- ReAct: Synergizing Reasoning and Acting in Language Models
- Learning to Keep a Promise: Scaling Language Model Decoding Parallelism with Learned Asynchronous Decoding
- Learning Adaptive Parallel Reasoning with Language Models
- ASPD: Unlocking Adaptive Serial-Parallel Decoding by Exploring Intrinsic Parallelism in LLMs
- ThreadWeaver: Adaptive Threading for Efficient Parallel Reasoning in Language Models
- A Survey on Parallel Text Generation: From Parallel Decoding to Diffusion Language Models
- SPRINT: Enabling Interleaved Planning and Parallelized Execution in Reasoning Models
- Qwen2.5 Technical Report
- Qwen3 Technical Report
- Large Language Monkeys: Scaling Inference Compute with Repeated Sampling
- Efficient Beam Search for Large Language Models Using Trie-Based Decoding
- Self-Consistency Improves Chain of Thought Reasoning in Language Models
- Universal Self-Consistency for Large Language Model Generation
- Accessing GPT-4 level Mathematical Olympiad Solutions via Monte Carlo Tree Self-refine with LLaMa-3 8B
- Hogwild! Inference: Parallel LLM Generation via Concurrent Attention
- Parallel-R1: Towards Parallel Thinking via Reinforcement Learning
- FairBatching: Fairness-Aware Batch Formation for LLM Inference
- TurboSpec: Closed-loop Speculation Control System for Optimizing LLM Serving Goodput
- AdaServe: Accelerating Multi-SLO LLM Serving with SLO-Customized Speculative Decoding
Related papers
- Programmable Probabilistic Computer with 1,000,000 p-bits
- BatchGen: An Architecture for Scalable and Efficient Batch Inference
- Seamless Reconfiguration for DAG-BFT (Technical Report)
- ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving
- iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems
- TIRx: A Unified Tile Primitive ML Compiler Across Heterogeneous Backends