ArA-DF 2026

ArabicNLP 2026 Shared Task

ArA-DF 2026

Arabic speech deepfake detection across dialect and robustness conditions. Evaluation and system-paper reviews are complete, gold labels are public, and camera-ready papers are due September 10.

Open Dataset Open Reviews Google Group
Task
Binary audio classification
Metric
Equal Error Rate
Venue
ArabicNLP 2026 @ EMNLP

Challenge evaluation and system-paper reviews are complete.

  • Final leaderboards
  • Gold labels released
  • Reviews released Sep 5
  • Camera-ready due Sep 10

Overview

A focused benchmark for Arabic audio authenticity.

ArA-DF 2026 is listed by ArabicNLP 2026 as Shared Task 14. Participating systems assign each Arabic speech sample to one of two labels: bona fide, for genuine human speech, or spoofed, for synthetic speech produced by text-to-speech or voice-conversion systems.

The task covers Modern Standard Arabic and Arabic dialects, with evaluation designed to reward detectors that generalize beyond narrow speakers, synthesis systems, and recording conditions.

The frozen final leaderboards include 14 ranked teams on Track 1 and 12 ranked teams on Track 2.

Tracks

Final leaderboards for two tracks.

Both tracks use Equal Error Rate as the primary ranking metric, with Accuracy and macro-F1 reported for additional interpretation.

Track 1

Dialect Generalization

Evaluate whether systems remain reliable on Arabic dialects and speakers that are not represented in the training data.

View final leaderboard Track 2

Acoustic Robustness

Test robustness under practical audio conditions such as compression artifacts, background noise, and re-recording effects.

View final leaderboard

Data & Baseline

Live dataset and baseline resources.

The ArA-DF 2026 dataset and baseline are hosted by ArabicSpeech on Hugging Face. Full loading commands and repository instructions remain on the Hugging Face cards.

Dataset on Hugging Face

Train/dev labels plus post-challenge gold labels for development-test and final-test configs.

Open dataset

Baseline on Hugging Face

wav2vec2 XLS-R 300M frontend with an AASIST backend and helper scripts.

Open baseline

Codabench leaderboards

Frozen final leaderboards and archived competition pages for both tracks.

Track 1 Track 2

ArabicNLP shared-tasks page

Conference-wide dates, contact listing, and paper-submission updates.

Open page

Dataset snapshot

Labels use 0 for spoofed or synthetic speech and 1 for bona fide speech. Released audio is 16 kHz mono PCM, packaged as lossless FLAC in WebDataset TAR shards, with metadata available through Hugging Face Parquet configs.

Dataset split rows and label availability
Split Rows Labels
Train22,500Yes
Dev21,000Yes
Track 1 development-test16,023Yes, in labeled config
Track 1 final-test144,210Yes, in labeled config
Track 2 development-test14,193Yes, in labeled config
Track 2 final-test127,746Yes, in labeled config

Baseline snapshot

The baseline repository includes a trained checkpoint, XLS-R frontend weights, configuration, parquet-generation utility, and published result file. Lower EER is better.

Baseline model-card EER results
Baseline split EER (%) Utterances
Track 1 development-test14.6716,023
Track 1 final-test14.54144,210
Track 2 development-test27.2114,193
Track 2 final-test27.11127,746

Values are baseline model-card results, not participant rankings.

Released August 4, 2026

Post-challenge test labels are now available.

The original challenge-era test configs remain blind for reproducibility. Gold labels are provided in separate Hugging Face configs whose names end in _labeled. The WebDataset TAR shards and JSON sidecars are unchanged and remain label-free.

Dataset card Loading examples
Released gold-label counts
Labeled config Rows Label 0 Label 1
track-1_development_test_labeled16,02311,9264,097
track-1_test_labeled144,210107,30736,903
track-2_development_test_labeled14,19310,6833,510
track-2_test_labeled127,74696,25631,490

System Papers

Reviews are available; camera-ready papers are next.

Reviews and acceptance notifications were released on September 5. Authors should sign in to OpenReview to read their feedback and prepare accepted papers for the final deadline.

Complete

System-paper submission

.

Released

Reviews and notifications

. Open venue.

Due next

Camera-ready deadline

, 11:59 PM UTC-12.

Required format

  • ArabicNLP/EMNLP short-paper format.
  • Maximum 4 pages, excluding unlimited references.
  • Non-anonymous submission.
  • Use the title format: <Team Name> at ArA-DF 2026: <Your Contribution>.
  • Use official ACL style files without modifying margins, fonts, or paper size.

What to include

  • System architecture, algorithms, and design decisions.
  • Data usage, preprocessing, training setup, tools, and versions.
  • Official submitted results and any clearly marked post-submission results.
  • Analysis, ablations, error patterns, limitations, and ethical considerations.
  • Code or model URL when available, to support reproducibility.

Recommended structure

  1. Abstract
  2. Introduction
  3. Background / Task Setup
  4. System Overview
  5. Experimental Setup
  6. Results and Analysis
  7. Conclusion
  8. Acknowledgments
  9. Appendix, if needed

Important Dates

Task timeline.

Completed challenge milestones and the remaining participant deadlines.

Resources released

Complete Training, dev and open test data, evaluation scripts, and baseline.

Development phase

Complete Experimentation and submissions on the development-test sets.

Evaluation phase

Complete Submissions on the final-test sets and official scoring.

Leaderboard freeze

Complete Final leaderboard state is frozen for official results.

Gold labels released

Complete Post-challenge test labels released through Hugging Face labeled configs.

System papers submitted

Complete Initial system-description paper deadline.

Reviews and notifications

Complete Reviews and acceptance notifications released through OpenReview.

Camera-ready due

Upcoming Camera-ready versions of accepted system papers due.

ArabicNLP 2026

Upcoming Hybrid conference alongside EMNLP 2026 in Budapest.

All deadlines are 11:59 PM UTC-12 unless otherwise stated. Reviews, decisions, and final-version controls are available through the ArA-DF OpenReview venue.

Evaluation

Ranked by Equal Error Rate.

Equal Error Rate is the primary metric because it is threshold-independent and widely used in audio anti-spoofing evaluation. Lower EER is better; official scores came from the blind final-test submissions during the evaluation phase.

EER Primary ranking metric
Accuracy Secondary leaderboard metric
Macro-F1 Secondary leaderboard metric

Contact & Forum

Where to ask questions and report issues.

Use the public Google Group for task questions, discussions, announcements, and clarifications. Route private, platform, and repository issues to the appropriate channel below.

System papers

Reviews and camera-ready versions

Sign in to the ArA-DF OpenReview venue to access reviews and decisions and manage the camera-ready version of an accepted paper.

OpenReview venue
Public forum

Questions, discussions, and announcements

Use the ArA-DF 2026 Google Group for participant-wide questions, official clarifications, system-paper questions, and shared announcements.

Open Google Group Email group
Private issues

Team-specific or sensitive support

Email the organizers for private issues involving authorship, reviews, decisions, or other sensitive concerns.

Email organizers
Final results

Codabench leaderboards

Use the archived competition pages to view the frozen final leaderboard for each track.

Track 1 Track 2
Dataset and baseline

Hugging Face repository issues

Use the Hugging Face dataset and baseline repositories for data, loading, baseline, and model-card questions.

Dataset Baseline

Organizers

Organizing team.

For public task questions, discussions, and announcements, use the ArA-DF 2026 Google Group: ara-df-2026@googlegroups.com.

Vasista Sai Lodagala

HUMAIN, Saudi Arabia

Yassine El Kheir

DFKI, Germany

Sara Althubaiti

HUMAIN, Saudi Arabia

Pedro Moreno Mengibar

HUMAIN, Saudi Arabia

Ahmed Ali

HUMAIN, Saudi Arabia