AudioTrust
Audio language models answer prompts that probe hallucination, robustness, speaker authentication, privacy, fairness, and safety. Automated evaluators, including language-model judges, compare responses with task-specific references and criteria.
The revised paper reports 26 subtasks and more than 4,420 audio samples. Metric definitions and directions differ across subtasks.
- Made by
- Nanyang Technological University, Tsinghua University, Beijing Normal-Hong Kong Baptist University and others
- Released
- May 2025
- Kind
- Audio understanding and voice
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Propensity
- Behavior
- Hallucination, Lawlessness, Bias, Harmful use
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Audio and speech, Content safety
- Strongest model tested
-
Gemini 2.5 Flash Sep 2025 #112 on the capability index
- Last model added
- September 2025
- Scores refreshed
- November 20th, 2026
In the authors’ words
The rapid development and widespread adoption of Audio Large Language Models (ALLMs) demand rigorous evaluation of their trustworthiness. However, existing evaluation frameworks are primarily designed for text and fail to capture vulnerabilities introduced by the acoustic properties of audio. We find that significant trustworthiness risks in ALLMs arise from non-semantic acoustic cues, such as timbre, accent, and background noise, which can be exploited to manipulate model behavior. To address this gap, we propose AudioTrust, the first large-scale and systematic framework for evaluating ALLM trustworthiness under audio-specific risks. AudioTrust covers six key dimensions: fairness, hallucination, safety, privacy, robustness, and authenticition. It includes 26 sub-tasks and a curated dataset of more than 4,420 audio samples collected from real-world scenarios, including daily conversations, emergency calls, and voice assistant interactions, and is specifically designed to probe trustworthiness across multiple dimensions. Our comprehensive evaluation spans 18 experimental settings and uses human-validated automated pipelines to enable objective and scalable assessment of model outputs. Experimental results on 14 state-of-the-art open-source and closed-source ALLMs reveal important limitations and failure boundaries under diverse high-risk audio scenarios, providing critical insights for the secure and trustworthy deployment of future audio models. Our platform and benchmark are publicly available at https://github.com/JusperLee/AudioTrust.
GPT-4o Audio tops the leaderboard with 0.93 Gemini 2.5 Flash leads with 0.37 among models released by September 2025
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. The paper gives no overall score, so the headline is its first metric, group fairness for social stereotypes. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Gemini 2.5 Flash, at 0.37. This view shows only models we track, each at its best.
The newest model with a score here came out in September 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 0.93 GPT-4o Audio
- 0.86 GPT-4o mini Audio
- 0.66 Step-Fun
- 0.37
Gemini 2.5 Flash
- 0.33
Qwen2-Audio-7B-Instruct
- 0.32 Gemini-2.5 Pro (snapshot unspecified)
- 0.30 Gemini-1.5 Pro (snapshot unspecified)
- 0.26 MiniCPM-o-2_6
- 0.24 Ultravox
- 0.14 SALMONN
- 0.07 Step-Audio2
- 0.07 Qwen2.5-Omni
- 0.04 Kimi-Audio-7B-Instruct
- 0.02 OpenS2S
- 1 GPT-4o Audio 0.93
- 2 GPT-4o mini Audio 0.86
- 3 Step-Fun 0.66
- 4
Gemini 2.5 Flash Sep 2025 0.37
- 5
Qwen2-Audio-7B-Instruct 0.33
- 6 Gemini-2.5 Pro (snapshot unspecified) 0.32
- 7 Gemini-1.5 Pro (snapshot unspecified) 0.30
- 8 MiniCPM-o-2_6 0.26
- 9 Ultravox 0.24
- 10 SALMONN 0.14
- 11 Step-Audio2 0.07
- 12 Qwen2.5-Omni 0.07
- 13 Kimi-Audio-7B-Instruct 0.04
- 14 OpenS2S 0.02
Source: The AudioTrust paper
- 1
Gemini 2.5 Flash Sep 2025 0.37
- 2
Qwen2-Audio-7B-Instruct 0.33
- 3 MiniCPM-o-2_6 0.26
- 4 Kimi-Audio-7B-Instruct 0.04
Source: The AudioTrust paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-4o Audio | 0.93 | Official leaderboard | ||
| 2 | GPT-4o mini Audio | 0.86 | Official leaderboard | ||
| 3 | Step-Fun | 0.66 | Official leaderboard | ||
| 4 | | 0.37 | Sep 2025 | Official leaderboard | |
| 5 | | 0.33 | Aug 2024 | Official leaderboard | |
| 6 | Gemini-2.5 Pro (snapshot unspecified) | 0.32 | Official leaderboard | ||
| 7 | Gemini-1.5 Pro (snapshot unspecified) | 0.30 | Official leaderboard | ||
| 8 | MiniCPM-o-2_6 | 0.26 | Jan 2025 | Official leaderboard | |
| 9 | Ultravox | 0.24 | Official leaderboard | ||
| 10 | SALMONN | 0.14 | Official leaderboard | ||
| 11 | Step-Audio2 | 0.07 | Official leaderboard | ||
| 12 | Qwen2.5-Omni | 0.07 | Official leaderboard | ||
| 13 | Kimi-Audio-7B-Instruct | 0.04 | Apr 2025 | Official leaderboard | |
| 14 | OpenS2S | 0.02 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 0.37 | Sep 2025 | Official leaderboard | |
| 2 | | 0.33 | Aug 2024 | Official leaderboard | |
| 3 | MiniCPM-o-2_6 | 0.26 | Jan 2025 | Official leaderboard | |
| 4 | Kimi-Audio-7B-Instruct | 0.04 | Apr 2025 | Official leaderboard |
Rankings by part
We read 2 parts from the benchmark’s paper: Social stereotypes and Decision-making. Each ranks models on one part of the test. Gemini 2.5 Flash leads Social stereotypes and MiniCPM-o-2_6 leads Decision-making. Arrows show how far a model moved from its overall rank among the models we track.
Social stereotypes
4 models we track have a score here.
- 1
Gemini 2.5 Flash Sep 2025 0.37
- 2
Qwen2-Audio-7B-Instruct 0.33
- 3 MiniCPM-o-2_6 0.26
- 4 Kimi-Audio-7B-Instruct 0.04
Decision-making
4 models we track have a score here.
- 1 MiniCPM-o-2_6 ↑2 0.41
- 2
Qwen2-Audio-7B-Instruct 0.29
- 3
Gemini 2.5 Flash Sep 2025 ↓2 0.25
- 4 Kimi-Audio-7B-Instruct 0.09
2 models have held number one in the benchmark’s paper since its release in May 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The AudioTrust paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official AudioTrust leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Audio and video
- Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy.
Qwen 3.8 Max 88.7%
- LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization.
Gemini 4 Argon 91.7%
- MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy.
Qwen 3.8 Max 82.4%
- MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy.
Qwen 3.8 Max 90.8%
- FLEURS In use Models transcribe read speech in 102 languages built from FLoRes-101 sentences. Model cards usually report word error rate, so lower scores are better.
Phi-4-multimodal 4.00%
- VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules.
Qwen3.5-Omni-Plus 93.1%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the benchmark’s paper (14 entries).
- Pass Models scored 4 models we track, plus 10 other entries on its leaderboard.
- Warning Newest score The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 2.5 Flash (#112).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 4 rows for models we track and 10 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.