Afrispeech Semantics - Consistency (AfriSpeech-General): leaderboard
Binary audio-text consistency on African-accented conversational speech from AfriSpeech-General (Afrispeech Semantics, Okocha and Grant, 2026): the model decides whether each human-verified hypothesis agrees or conflicts with the spoken premise, with no neutral option. Macro-F1 in percent for audio language models prompted zero-shot; higher is better.
Metric: Macro-F1 (%; binary consistent versus inconsistent judgments of text hypotheses against a spoken premise, computed over all premise-hypothesis pairs of the AfriSpeech-General set; zero-shot generative audio language model, free-form answers mapped to labels by a lightweight language model). Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen2.5-Omni-7B | 91.31 | #1797 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=afrispeech-semantics-consistency-afrispeech-general · How It Works · Data refreshed daily, snapshot 2026-11-12.