AT-ADD Challenge

Welcome to the AT-ADD Challenge at ACM Multimedia 2026

The rapid progress of audio generation models has made it possible to synthesize highly realistic speech, environmental sound, singing voice, and music, creating new challenges for multimedia authenticity and security.

The AT-ADD Challenge (All-Type Audio Deepfake Detection) aims to advance robust and generalizable detection methods for real-world and emerging audio deepfakes. It includes two tracks: Track 1 for robust speech deepfake detection, and Track 2 for all-type audio deepfake detection across speech, sound, singing, and music.

We welcome participants from academia and industry to join the challenge and help shape the future of trustworthy audio forensics.

Quick Access

Official Paper | Baseline Code

Registration for Track 1 | Registration for Track 2

Codabench for Track 1 | Codabench for Track 2

Challenge Overview

The AT-ADD challenge is designed as a two-track evaluation framework that systematically assesses both robustness and generalization in audio deepfake detection.

🔹 Track 1: Robust Speech Deepfake Detection
Focuses on evaluating the robustness of countermeasures under realistic conditions, including domain shifts, acoustic distortions, and unseen state-of-the-art deepfake methods.

🔹 Track 2: All-Type Audio Deepfake Detection
Extends detection beyond speech to heterogeneous audio types (speech, sound, singing voice, and music), emphasizing cross-type generalization and type-agnostic modeling.

Participants are expected to develop robust and generalizable countermeasures capable of detecting deepfake audio across diverse conditions, unseen generation methods, and multiple audio types.

Important Dates

Apr 8 ◆ Release of Data, Baseline Code, and Paper

▶ Progress Evaluation Stage opens ← Ongoing

Jun 8 Opening of Final Evaluation Stage

Jun 15 Final Leaderboard Freeze

Jun 17 Metadata & Technical Report Submission

Jun 25 Participant Paper Submission Deadline

Jul 16 Notification of Paper Acceptance

Aug 6 Camera-Ready Paper Submission Deadline

Three outstanding technical solution papers — Track 1 Top 1, Track 2 Top 1, and best solution — will be recommended for inclusion in the ACM MM 2026 main conference proceedings.

Get in Touch

Organization Inquiries: Haonan Cheng (haonancheng@cuc.edu.cn), Jiayi Zhou (zjy326112@antgroup.com)

Technical Questions: Yuankun Xie (xieyuankun@cuc.edu.cn), Tao Wang (mengyu.wt@antgroup.com)