Hugging Face announced the addition of Egyptian and Nigerian English accents to its Open ASR Leaderboard, marking the latest expansion of private evaluation data.confirmed According to the company's blog post, this move follows previous additions of private data from Appen, DataOceanAI, and Voice Arena.company claimHugging Face has added new accent data to its speech recognition leaderboard to help test how well AI understands different English varieties.Hugging Face integrated private Egyptian and Nigerian English splits from Liva AI into the Open ASR Leaderboard, enhancing held-out evaluation capabilities.
The claim
The company states that these new splits provide a richer breakdown of accents based on Kachru’s Three Circles model. The leaderboard will now distinguish between US versus non-US accents and Inner Circle versus Outer and Expanding Circles. Hugging Face credits Hume AI with the idea for this classification structure.
The response
Community feedback highlighted the importance of held-out sets for trustworthy evaluations, according to the published report The company notes that previous private data additions received strong community response, driving the decision to continue improving representation in ASR evaluations.reported
The practical upshot
As more accents join the leaderboard, Hugging Face plans to split the Outer and Expanding Circles further. The current update allows for reporting on US, UK, Australia, Canada, India, Nigeria, and Egypt, providing a broader view of global ASR performance.
Claims versus evidence
Hugging Face claims that using private data minimizes "benchmaxxing," or over-optimization for public benchmarks. The evidence rests on the use of held-out sets from Liva AI, which collects data through real consumer product usage rather than artificial elicitation. The source acknowledges that this approach provides natural, high-context behavior, but does not provide independent third-party validation of the reduction in benchmark gaming.
Why it matters
- Product teams can now evaluate ASR models against Egyptian and Nigerian English, addressing gaps in non-US accent coverage.
- The shift to private, held-out data sets may offer more realistic performance metrics than public benchmarks alone.
- Using Kachru’s Three Circles model provides a standardized framework for comparing inner versus outer circle accent performance.
- Liva AI’s method of collecting data from real product use suggests higher ecological validity for interaction-based tasks.
Topics: ASR, Hugging Face, Liva AI, benchmarks, speech recognition
Updated
Then / now / next
The Open ASR Leaderboard featured private data from Appen, DataOceanAI, and Voice Arena covering five accents.
Egyptian and Nigerian English accents have been added, expanding the leaderboard’s accent coverage and classification.
Hugging Face plans to split the Outer and Expanding Circles further as more accents join the leaderboard.
Story timeline
- Then The Open ASR Leaderboard featured private data from Appen, DataOceanAI, and Voice Arena covering five accents.
- 06 Nov 2026 Egyptian and Nigerian English join the Open ASR Leaderboard
- Now Egyptian and Nigerian English accents have been added, expanding the leaderboard’s accent coverage and classification.
- Next Hugging Face plans to split the Outer and Expanding Circles further as more accents join the leaderboard.
Ask the article
Local keyword search of this article; no generated answers.Which of these is confirmed?
Check the answer
Hugging Face announced the addition of Egyptian and Nigerian English accents to its Open ASR Leaderboard, marking the latest expansion of private evaluation data.
The matched primary source and announcement wording support the confirmed label.
Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
ASR
Automatic Speech Recognition, the technology that converts spoken language into text.
Benchmaxxing
Over-optimizing a model’s performance on a specific public benchmark, potentially reducing real-world generalization.
Kachru’s Three Circles
A model classifying varieties of English into Inner, Outer, and Expanding circles based on historical and social factors.
Held-out sets
Data sets reserved for evaluation that are not used during model training, ensuring unbiased performance measurement.
Questions readers ask
What new accents are added to the Open ASR Leaderboard?
Egyptian and Nigerian English accents have been added via private data splits from Liva AI.
How does Hugging Face classify accents on the leaderboard?
The leaderboard uses Kachru’s Three Circles model, distinguishing between US/non-US and Inner/Outer & Expanding Circles.
Where does the new data come from?
The data is collected by Liva AI through consumer products used by real people, providing natural interaction data.
Connected stories
Reading receipt
- Hugging Face integrated two new private accent splits: Egyptian and Nigerian English.
- The data comes from Liva AI, which collects interaction data via consumer products.
- The leaderboard now reports metrics using Kachru’s Three Circles model for accents.
Still open: How will the future splitting of Outer and Expanding Circles affect benchmark comparability?
Still open: What specific performance differences emerge between Inner and Outer Circle accents on these new sets?