Rethinking in Spikes: Mitigating Hallucinations in MDLMs with Step-Aware Decoding

媒体报道 Hugging Face Papers 原文语言:中文

论文摘要(作者报告,未经本站复测):Recent advancements in multimodal diffusion language models (MDLMs) have exhibited strong global modeling capabilities through iterative refinement.

全文

StepRefine, a plug-and-play decoding strategy, mitigates hallucinations in multimodal diffusion language models by switching between discrete denoising and latent refinement based on entropy spikes and distributional stability.

正文以来源语言显示。

参考来源

← 返回每日AI报告