Weaver builds conditional proposal trees from a factorized drafter’s top-K marginals and a rollback-free GDN tree-verify kernel, yielding 4.37× speedup over AR decoding.
Accelerating llm inference with staged spec- ulative decoding
12 Pith papers cite this work. Polarity classification is still indexing.
12
Pith papers citing it
abstract
Recent advances with large language models (LLM) illustrate their diverse capabilities. We propose a novel algorithm, staged speculative decoding, to accelerate LLM inference in small-batch, on-device scenarios. We address the low arithmetic intensity of small-batch inference by improving upon previous work in speculative decoding. First, we restructure the speculative batch as a tree, which reduces generation costs and increases the expected tokens per batch. Second, we add a second stage of speculative decoding. Taken together, we reduce single-batch decoding latency by 3.16x with a 762M parameter GPT-2-L model while perfectly preserving output quality.
citation-role summary
background 3
citation-polarity summary
roles
background 3polarities
background 3representative citing papers
CaDDTree jointly selects tree structure and budget to maximize expected tokens per unit time in speculative decoding, proving unimodality under convex verification cost and matching oracle DDTree performance on Qwen models.
Establishes the first rigorous framework for continuous semantic caching of LLM responses using ε-net discretization and kernel ridge regression, with sublinear regret bounds.
WISV uses a channel-aware semantic acceptance policy on hidden representations to boost accepted sequence length by up to 60.8% and cut interaction rounds by 37.3% in distributed speculative decoding, with under 1% accuracy loss.
Medusa augments LLMs with multiple decoding heads and tree-based attention to predict and verify several tokens in parallel, yielding 2.2-3.6x inference speedup via two fine-tuning regimes.
StickyInvoc introduces sticky tasks that load LLM model state once and invocation tasks that reuse it, yielding 3.6x speedup on a 150k-inference workflow.
CLP is a lightweight linear predictor for safe multi-token spans in LLM decoding that delivers 1.14x-1.29x speedup on Qwen2.5 models with zero measured quality degradation.
SV-VLA uses infrequent heavy VLA planning of action chunks plus a lightweight closed-loop verifier to achieve both efficiency and robustness in dynamic robot control.
EAGLE resolves feature-level uncertainty in speculative sampling via one-step token advancement, delivering 2.7x-3.5x speedup on LLaMA2-Chat 70B and doubled throughput across multiple model families and tasks.
RTP-LLM is a new LLM inference engine achieving 4.7x-6.3x model loading speedup and 1.12x-2.52x throughput gains over vLLM and SGLang via disaggregated phases, multi-tier KV cache, and modular optimizations in production at Alibaba.
The paper surveys techniques to speed up and reduce the resource needs of LLM inference, organized by data-level, model-level, and system-level changes, with comparative experiments on representative methods.
A literature survey of Small Language Models (1-8B parameters) that can perform comparably or better than larger models, covering general-purpose and task-specific approaches plus creation techniques.
citing papers explorer
-
Trees from Marginals: Autoregressive drafting with factorized priors
Weaver builds conditional proposal trees from a factorized drafter’s top-K marginals and a rollback-free GDN tree-verify kernel, yielding 4.37× speedup over AR decoding.
-
Cost-Aware Diffusion Draft Trees for Speculative Decoding
CaDDTree jointly selects tree structure and budget to maximize expected tokens per unit time in speculative decoding, proving unimodality under convex verification cost and matching oracle DDTree performance on Qwen models.
-
Continuous Semantic Caching for Low-Cost LLM Serving
Establishes the first rigorous framework for continuous semantic caching of LLM responses using ε-net discretization and kernel ridge regression, with sublinear regret bounds.
-
WISV: Wireless-Informed Semantic Verification for Distributed Speculative Decoding in Device-Edge LLM Inference
WISV uses a channel-aware semantic acceptance policy on hidden representations to boost accepted sequence length by up to 60.8% and cut interaction rounds by 37.3% in distributed speculative decoding, with under 1% accuracy loss.
-
Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads
Medusa augments LLMs with multiple decoding heads and tree-based attention to predict and verify several tokens in parallel, yielding 2.2-3.6x inference speedup via two fine-tuning regimes.
-
StickyInvoc: Rethinking Task Models for High-throughput Workflows in the LLM Era
StickyInvoc introduces sticky tasks that load LLM model state once and invocation tasks that reuse it, yielding 3.6x speedup on a 150k-inference workflow.
-
CLP: Collocation-Length Prediction for Zero-Loss Adaptive Multi-Token Inference
CLP is a lightweight linear predictor for safe multi-token spans in LLM decoding that delivers 1.14x-1.29x speedup on Qwen2.5 models with zero measured quality degradation.
-
Open-Loop Planning, Closed-Loop Verification: Speculative Verification for VLA
SV-VLA uses infrequent heavy VLA planning of action chunks plus a lightweight closed-loop verifier to achieve both efficiency and robustness in dynamic robot control.
-
EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty
EAGLE resolves feature-level uncertainty in speculative sampling via one-step token advancement, delivering 2.7x-3.5x speedup on LLaMA2-Chat 70B and doubled throughput across multiple model families and tasks.
-
RTP-LLM: High-Performance Alibaba LLM Inference Engine
RTP-LLM is a new LLM inference engine achieving 4.7x-6.3x model loading speedup and 1.12x-2.52x throughput gains over vLLM and SGLang via disaggregated phases, multi-tier KV cache, and modular optimizations in production at Alibaba.
-
A Survey on Efficient Inference for Large Language Models
The paper surveys techniques to speed up and reduce the resource needs of LLM inference, organized by data-level, model-level, and system-level changes, with comparative experiments on representative methods.
-
Small Language Models (SLMs) Can Still Pack a Punch: A survey (updated 2026)
A literature survey of Small Language Models (1-8B parameters) that can perform comparably or better than larger models, covering general-purpose and task-specific approaches plus creation techniques.