AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Beyond Scattered Acceptance: Fast and Coherent Inference for DLMs via Longest Stable Prefixes

Pengxiang Li, Jiayin Cai, Hongwei Xue, Kunyu Shi, Shilin Yan

ICLR 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
use adaptive prefix sizing instead of fixed 8-token commits (ablation)LSP · GSM8K (LLaDA-8B)score (%): 19.3 -> 69.9Improved
use adaptive sizing instead of fixed 1-token commits (ablation)LSP · GSM8K (LLaDA-8B)total steps: 128 -> ~68 steps, 67.1 -> 69.9 scoreImproved
add structural snapping to delimiters (ablation)LSP · GSM8K (LLaDA-8B)score (%): 67.8 -> 69.9 (steps ~50 -> ~68)Improved
use prefix-first instead of scattered acceptance (ablation)LSP · GSM8K (LLaDA-8B)score (%): 68.9 -> 69.9 (steps 128 -> ~68)Improved
use prefix-first instead of scattered acceptance (ablation)LSP · mid-generation flip rateflip rate (%): 14.2% -> 4.3%Improved
apply LSP scheduler vs Full decodingLLaDA-8B · MMLU / HumanEval / GSM8Kscore (%) and speedup: GSM8K 77.1 -> 77.6 at 1.51x; HumanEval 30.5 -> 29.3 at 1.22xNo clear effect
apply LSP scheduler vs Full decodingLLaDA-8B · TruthfulQAscore (%): 34.4 -> 45.8 (2.29x)Improved
apply LSP scheduler vs Full decodingLLaDA-8B · WinoGrandescore (%): 73.8 -> 70.7 (1.74x)Hurt
apply LSP scheduler vs Full decodingDream-7B · Sudokuspeedup: 3.36x speedup, 89.0 -> 88.0Improved
apply LSP scheduler vs Full decodingLLaDA-8B · WritingPromptscoherence (1-5): 4.42 -> 4.38 at 1.82xNo 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