AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Swordsman: Entropy-Driven Adaptive Block Partition for Efficient Diffusion Language Models

Yu Zhang, Xinchen Li, Jialei Zhou, Hongnan Ma, Zhongwei Wan, Yiwei Shi, Duoqian Miao, Qi Zhang, Longbing Cao

ICML 2026 · Image generation › Fast diffusion sampling · approach: Masked diffusion language model decoding · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace fixed-size blocks with entropy-driven adaptive block partition (ablation)Swordsman on LLaDA-8B-Instruct · GSM8K (no cache)accuracy: 77.56 -> 81.43Improved
add block-wise decoding to full-diffusion LLaDA (ablation)LLaDA-8B-Instruct · GSM8Kaccuracy: 29.04 -> 77.40Improved
add confidence-based parallel decoding to block-wise decoding (ablation)LLaDA-8B-Instruct · GSM8Kthroughput (TPS): 8.39 -> 42.36Improved
use dynamic threshold (tau_init 0.9) instead of static threshold 0.9 (ablation)Swordsman on LLaDA-8B-Instruct · GSM8K (no cache)latency (s): accuracy 81.45 vs 81.43; latency reduced by 2.49sImproved
lower initial dynamic threshold to 0.5 (ablation)Swordsman on LLaDA-8B-Instruct · GSM8K (no cache)accuracy: 81.43 -> 68.46Hurt
raise initial dynamic threshold to 1.0 (ablation)Swordsman on LLaDA-8B-Instruct · GSM8K (no cache)latency (s): about 2.4x higher latencyHurt
use Swordsman instead of Fast-dLLM under dual cacheLLaDA-1.5 · HumanEval (dual cache)accuracy: 35.59 -> 43.90Improved
use Swordsman instead of Fast-dLLM under dual cacheLLaDA-8B-Instruct · GSM8K (dual cache)accuracy: 75.21 -> 81.50Improved
use Swordsman instead of Fast-dLLM under dual cacheLLaDA-8B-Instruct · GSM8K speedthroughput (TPS): 72.12 -> 73.74Improved
use Swordsman instead of Fast-dLLM on MATH under dual cacheLLaDA-1.5 · MATH (dual cache)accuracy: 33.26 -> 35.30Improved
use Swordsman instead of Fast-dLLM on MBPP under dual cacheLLaDA-8B-Instruct · MBPP (dual cache)accuracy: 13.60 -> 13.60No clear effect
use Swordsman instead of Fast-dLLM on MATH under dual cacheDream-v0-Base-7B · MATH (dual cache)accuracy: 38.34 -> 38.58No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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