Papers & Research

Tencent Hunyuan and partners release MMAE, the first audio editing benchmark

Original title:Can AI truly edit audio, not just generate it? 🎧 Tencent Hy, in collaboration with SJTU, SII, NTU,...

AI summary

Tencent Hunyuan, together with Shanghai Jiao Tong University, Nanyang Technological University, and other institutions, has launched MMAE (Massive Multitask Audio Editing Benchmark), the first benchmark to comprehensively evaluate AI speech/audio editing capabilities. MMAE requires models to understand existing audio and precisely modify it according to natural language instructions rather than simply generate it. Current models score below 5% exact match rate (EMR) on the benchmark, exposing shortcomings in reliable audio editing. MMAE contains 2,000 high-fidelity samples from real scenarios and 17,741 fine-grained evaluation items, covering 7 modalities across sound, music, speech, and mixtures, 6 task complexity levels (from basic modification to multi-hop reasoning and multi-turn editing), and 8 operation types (from local to global). The paper, code, dataset, and demo have been made public.

Key Takeaways

  • Tencent Hunyuan,together with Shanghai Jiao Tong University,Nanyang Technological University,and…
  • MMAE requires models to understand existing audio and precisely modify it according to natural…
  • Current models score below 5% exact match rate (EMR) on the benchmark,exposing shortcomings in…
  • MMAE contains 2,000 high-fidelity samples from real scenarios and 17,741 fine-grained evaluation…
  • The paper,code,dataset,and demo have been made public.

Quick Facts

Published
Updated
Source
X:腾讯混元 (@TencentHunyuan)
Category
Papers & Research
AI score
69
Read the original Read the full article at the original source
Why it matters

过去一年语音和音乐生成很热,但音频编辑还没人正经测过,腾讯这个基准把现状血淋淋地摆出来了,不到5%的准确率意味着整个方向都还在起步期。

Related entities

Source & copyright

This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.

Original article · AIHOT entry

How to Cite

When citing this article, keep the title, date, and permalink, and credit Orange News as the source.

Orange News: Tencent Hunyuan and partners release MMAE, the first audio editing benchmark. June 8, 2026. https://new.nat6.net/en/news/cmq4tfme001jlslt2hle9mu5a/

View all →