AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

LEAP: Zone-Aware MCTS for LLM Self-Speculative Decoding

LeiQuan Zheng, Yuan Liu

ICML 2026 · Large language models › Reasoning and chain of thought · approach: Speculative decoding

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes LEAP, a plug-and-play method that frames draft-model construction for LLM self-speculative decoding as a layer-configuration search solved with zone-aware Monte Carlo Tree Search.

Problem

Self-speculative decoding builds a draft model from a subset of the target LLM's layers, and choosing which layers to keep (the layer configuration) is the key challenge for speed-up.

Approach

Formulates draft model construction as an optimization problem over layer configurations and solves it with zone-aware Monte Carlo Tree Search, as stated in the abstract; the specific zone mechanism beyond that is not described in the available text.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add zone partitioning (ablation)LEAP · Overall (GSM8K, MT-Bench, WMT14), LLaMA-3-8Bspeedup: 1.67× -> 2.10×Improved
add layer grouping (ablation)LEAP · Overall (GSM8K, MT-Bench, WMT14), LLaMA-3-8Bspeedup: 1.63× -> 2.10×Improved
use MCTS instead of random search (ablation)LEAP · Overall (GSM8K, MT-Bench, WMT14), LLaMA-3-8Bspeedup: 1.56× -> 2.10×Improved
add zone partitioning (ablation)LEAP · MT-Bench, LLaMA-3-8Bspeedup: 1.64× -> 2.07×Improved
add layer grouping (ablation)LEAP · MT-Bench, LLaMA-3-8Bspeedup: 1.48× -> 2.07×Improved
add layer grouping (ablation)LEAP · GSM8K, LLaMA-3-8Bspeedup: 1.89× -> 2.12×Improved
use MCTS instead of random search (ablation)LEAP · WMT14, LLaMA-3-8Bspeedup: 1.51× -> 2.11×Improved
raise sampling temperature from 0 to 1LEAP · Overall 5 datasets, LLaMA-3-8Bspeedup: 2.05× -> 1.68×Hurt
raise sampling temperature from 0 to 1SWIFT · Overall 5 datasets, LLaMA-3-8Bspeedup: 1.13× -> 0.76×Hurt
raise sampling temperature from 0 to 1CLaSP · Overall 5 datasets, LLaMA-3-8Bspeedup: 1.27× -> 1.06×Hurt
scale target model from 8B to 70BSWIFT · Overall 5 datasets, T=0speedup: 1.13× -> 1.41×Improved
scale target model from 8B to 70BLEAP · Overall 5 datasets, T=0speedup: 2.05× -> 1.94×No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy