AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Stop the Flip-Flop: Context-Preserving Verification for Fast Revocable Diffusion Decoding

Yanzheng Xiang, Lan Wei, Yizhen Yao, Qinglin Zhu, Hanqi Yan, Chen Jin, Philip Alexander Teare, Dandan Zhang, Lin Gui, Amrutha Saseendran, Yulan He

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
add KV cache override with diagonal correction (ablation)COVER · GSM8K (LLaDA-Ins-8B, len 256)decoding steps: 123.28 -> 51.65Improved
add KV cache override with diagonal correction (ablation)COVER · GSM8K (LLaDA-Ins-8B, len 256)relative speed: 0.63x -> 1.00xImproved
add KV cache override with diagonal correction (ablation)COVER · HumanEval (LLaDA-Ins-8B, len 256)accuracy: 39.63 -> 40.24Improved
add KV cache override with diagonal correction (ablation)COVER · MATH500 (LLaDA-Ins-8B, len 256)accuracy: 30.60 -> 32.80Improved
add stability-aware seed selection (ablation)COVER · GSM8K (LLaDA-Ins-8B, len 256)decoding steps: 65.10 -> 51.65Improved
add stability-aware seed selection (ablation)COVER · HumanEval (LLaDA-Ins-8B, len 256)decoding steps: 102.14 -> 96.16Improved
add stability-aware seed selection (ablation)COVER · HumanEval (LLaDA-Ins-8B, len 256)accuracy: 39.02 -> 40.24Improved
use COVER instead of greedy baselineLLaDA-Ins-8B · GSM8K (len 512)relative speed: 1.00x -> 5.52xImproved
use COVER instead of greedy baselineDream-Ins-7B · MBPP (len 512)relative speed: 1.00x -> 11.64xImproved
use COVER instead of greedy baselineLLaDA-Base-8B · HumanEval (len 512)accuracy: 32.93 -> 35.37Improved
use COVER instead of WINOLLaDA-Ins-8B · MBPP (len 256)accuracy: 36.20 -> 39.00Improved
add Fast-dLLM prefix cacheCOVER · GSM8K (LLaDA-Ins-8B)relative speed: 1.00x -> 1.23xImproved

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