# Paralinguistic speech understanding on AIR-Bench Speech (test) Source: [Paralinguistic speech understanding on AIR-Bench Speech (test)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:41:58.346Z Evaluation of model performance on the test split of the AIR-Bench Speech dataset, focusing on paralinguistic understanding tasks including Gender, Emotion, Age, LID, Entity, and Intent recognition. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Paralinguistic speech understanding | | Dataset (including any stated split) | AIR-Bench Speech (test) | | Metric | Emotion Acc | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 11 of 11 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Emotion Acc](): Higher is better; unit: PERCENT. - [Gender Acc](): Higher is better; unit: PERCENT. - [Age Acc](): Higher is better; unit: PERCENT. - [LID Acc](): Higher is better; unit: PERCENT. - [Entity Acc](): Higher is better; unit: PERCENT. - [Intent Acc](): Higher is better; unit: PERCENT. - [Average Accuracy](): Higher is better; unit: PERCENT. - [Chat Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | Emotion Acc | Result date | Report paper | | --- | --- | --- | --- | | AZEROS | 71.45 | Not recorded | 6965d1596343fa78c83232db | | DeSTA2.5 | 64.3 | Not recorded | 6965d1596343fa78c83232db | | Qwen3-Omni-30B | 62.2 | Not recorded | 6965d1596343fa78c83232db | | Gemini2.5-pro | 60.7 | Not recorded | 6965d1596343fa78c83232db | | Whisper + GPT-4o | 59.5 | Not recorded | 6965d1596343fa78c83232db | | Qwen2.5-Omni | 54.85 | Not recorded | 6965d1596343fa78c83232db | | Whisper + Qwen2.5 | 50.8 | Not recorded | 6965d1596343fa78c83232db | | GPT-4o | 49.1 | Not recorded | 6965d1596343fa78c83232db | | Qwen2-Audio | 48.15 | Not recorded | 6965d1596343fa78c83232db | | SALMONN | 29.9 | Not recorded | 6965d1596343fa78c83232db | | GLM-4-Voice | 22.95 | Not recorded | 6965d1596343fa78c83232db | ### AZEROS Emotion Acc: 71.45. Result ID: 6996a92a5698665ec4f307e8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Paradigm":"End-to-end Speech-LLM"}. This record does not assert independent reproduction. ### DeSTA2.5 Emotion Acc: 64.3. Result ID: 6996a92a5698665ec4f307e5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Paradigm":"End-to-end Speech-LLM"}. This record does not assert independent reproduction. ### Qwen3-Omni-30B Emotion Acc: 62.2. Result ID: 6996a92a5698665ec4f307e7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Paradigm":"End-to-end Speech-LLM"}. This record does not assert independent reproduction. ### Gemini2.5-pro Emotion Acc: 60.7. Result ID: 6996a92a5698665ec4f307e1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Paradigm":"End-to-end Speech-LLM"}. This record does not assert independent reproduction. ### Whisper + GPT-4o Emotion Acc: 59.5. Result ID: 6996a92a5698665ec4f307de. [Reporting paper]() Recorded evaluation configuration: {"System Paradigm":"Cascaded System","ASR Module":"Whisper-large-v3","LLM Backbone":"GPT-4o"}. This record does not assert independent reproduction. ### Qwen2.5-Omni Emotion Acc: 54.85. Result ID: 6996a92a5698665ec4f307e6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Paradigm":"End-to-end Speech-LLM"}. This record does not assert independent reproduction. ### Whisper + Qwen2.5 Emotion Acc: 50.8. Result ID: 6996a92a5698665ec4f307df. [Reporting paper]() Recorded evaluation configuration: {"System Paradigm":"Cascaded System","ASR Module":"Whisper-large-v3","LLM Backbone":"Qwen2.5-7B-Instruct"}. This record does not assert independent reproduction. ### GPT-4o Emotion Acc: 49.1. Result ID: 6996a92a5698665ec4f307e0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Paradigm":"End-to-end Speech-LLM"}. This record does not assert independent reproduction. ### Qwen2-Audio Emotion Acc: 48.15. Result ID: 6996a92a5698665ec4f307e4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Paradigm":"End-to-end Speech-LLM"}. This record does not assert independent reproduction. ### SALMONN Emotion Acc: 29.9. Result ID: 6996a92a5698665ec4f307e2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"System Paradigm":"End-to-end Speech-LLM"}. This record does not assert independent reproduction. ### GLM-4-Voice Emotion Acc: 22.95. Result ID: 6996a92a5698665ec4f307e3. [Reporting paper]() Recorded evaluation configuration: {"System Paradigm":"End-to-end Speech-LLM"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.