研究人员推出了MMAE，一个旨在评估基于指令的音频编辑能力的全新基准。该基准涵盖七种音频模态，并包含从基本编辑到多步推理等不同复杂度的任务。对当前领先模型的评估显示，它们面临巨大挑战，精确匹配率低于5%，在复杂、混合模态任务上更是降至0%，这凸显了在精确执行和结构鲁棒性方面存在的关键局限性。 AI
影响 凸显了当前AI模型在音频编辑方面存在的显著差距，可能指导未来智能创作工具的发展。
排序理由 该集群包含一篇介绍AI能力评估新基准的研究论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
arXiv:2606.07229v1 Announce Type: cross Abstract: We introduce MMAE, a Massive Multitask Audio Editing benchmark, serving as the first comprehensive evaluation testbed designed for general-purpose instruction-based audio editing. Spurred by the shift toward intelligent creation, …
We introduce MMAE, a Massive Multitask Audio Editing benchmark, serving as the first comprehensive evaluation testbed designed for general-purpose instruction-based audio editing. Spurred by the shift toward intelligent creation, interactive editing has rapidly expanded from visu…
MMAE presents a comprehensive benchmark for instruction-based audio editing across multiple modalities and complexity levels, revealing significant gaps in current model capabilities.