# 🎙️ VoxSafeBench
**Not Just *What* Is Said, but *Who*, *How*, and *Where***
[](https://amphionteam.github.io/VoxSafeBench_demopage/)
[](https://arxiv.org/abs/2604.14548)
[](https://huggingface.co/datasets/YuxiangW/VoxSafeBench)
---
**VoxSafeBench** is a comprehensive benchmark designed to assess the social alignment of Speech Language Models (SLMs) built around three core pillars: **Safety**, **Fairness**, and **Privacy**. VoxSafeBench adopts a unique Two-Tier design: Tier 1 evaluates content-centric risks with matched text and audio inputs, while Tier 2 evaluates audio-conditioned risks in which the transcript is benign but the correct response depends on who is speaking, how they speak, or where they speak.
---
📑 Table of Contents
- [🎙️ VoxSafeBench](#️-voxsafebench)
- [⚙️ Environment Setup](#️-environment-setup)
- [📥 Dataset download](#-dataset-download)
- [🤖 Models download](#-models-download)
- [Qwen3-Omni](#qwen3-omni)
- [Qwen3-Omni-thinking](#qwen3-omni-thinking)
- [Mimo-audio/Mimo-audio-thinking](#mimo-audiomimo-audio-thinking)
- [Kimi-audio](#kimi-audio)
- [Gemini-3 \& GPT-4o-audio](#gemini-3--gpt-4o-audio)
- [🚀 Model Inference (Unified Runner)](#-model-inference-unified-runner)
- [Basic Usage](#basic-usage)
- [Advanced Options](#advanced-options)
- [Output Location](#output-location)
- [📈 Final Evaluation](#-final-evaluation)
- [Prerequisite](#prerequisite)
- [Basic Usage](#basic-usage-1)
- [Advanced Options](#advanced-options-1)
- [Output Location](#output-location-1)
- [⚠️ Note](#️-note)
## ⚙️ Environment Setup
First, clone the repository and navigate into the directory:
```bash
git clone https://github.com/AmphionTeam/VoxSafeBench.git
cd VoxSafeBench
```
Evaluating different models requires configuring their respective environments under each model's code repository.
For **closed-source models** (e.g., `gemini_3_flash`, `gemini_3_pro` and `gpt_4o_audio`) and the overall evaluation scripts (`run_inference.py`, `run_evaluation.py`), you need to install the following dependencies:
```bash
pip install openai google-genai python-dotenv tqdm
```
For **open-source models**, please refer to the specific setup instructions in their respective official repositories:
- **Qwen3-omni:** [QwenLM/Qwen3-Omni](https://github.com/QwenLM/Qwen3-Omni)
- **Mimo-audio:** [XiaomiMiMo/MiMo-Audio](https://github.com/XiaomiMiMo/MiMo-Audio)
- **Kimi-audio:** [MoonshotAI/Kimi-Audio](https://github.com/MoonshotAI/Kimi-Audio)
---
## 📥 Dataset download
```bash
export HF_ENDPOINT=https://hf-mirror.com
HF_HUB_DOWNLOAD_TIMEOUT=240 huggingface-cli download --repo-type dataset --resume-download YuxiangW/VoxSafeBench --local-dir ./datasets --max-workers 32
```
If the error is caused by network issues, you can try a few more times.
## 🤖 Models download
### Qwen3-Omni
```bash
export HF_ENDPOINT=https://hf-mirror.com
huggingface-cli download --resume-download Qwen/Qwen3-Omni-30B-A3B-Instruct --local-dir ./model_warehouse/Qwen3_omni
```
### Qwen3-Omni-thinking
```bash
export HF_ENDPOINT=https://hf-mirror.com
huggingface-cli download --resume-download Qwen/Qwen3-Omni-30B-A3B-Thinking --local-dir ./model_warehouse/Qwen3_omni_thinking
```
### Mimo-audio/Mimo-audio-thinking
```bash
export HF_ENDPOINT=https://hf-mirror.com
huggingface-cli download --resume-download XiaomiMiMo/MiMo-Audio-7B-Instruct --local-dir ./model_warehouse/Mimo_audio/MiMo-Audio-7B-Instruct
huggingface-cli download --resume-download XiaomiMiMo/MiMo-Audio-Tokenizer --local-dir ./model_warehouse/Mimo_audio/MiMo-Audio-Tokenizer
```
### Kimi-audio
```bash
export HF_ENDPOINT=https://hf-mirror.com
huggingface-cli download --resume-download moonshotai/Kimi-Audio-7B-Instruct --local-dir ./model_warehouse/Kimi_audio
```
> ⚠️ **Please replace** `model_warehouse/Kimi_audio/modeling_moonshot_kimia.py` with `utils/modeling_moonshot_kimia.py`.
> ⚠️ If you don't make this replacement, Kimi-audio cannot perform pure text input.
### Gemini-3 & GPT-4o-audio
Please put your `gemini_key` and `Openai_key` in the `.env` file.
---
## 🚀 Model Inference (Unified Runner)
We provide a unified inference script `run_inference.py` to generate responses from any model across various tasks seamlessly.
### Basic Usage
You can use the unified runner by specifying the model and the task:
```bash
# Run a specific task for a model
python run_inference.py --model Qwen3_omni --task Safety-tier2/Emotion
# Run all tasks for a model
python run_inference.py --model Qwen3_omni --all
```
**Available Models:**
- `Qwen3_omni`
- `Qwen3_omni_thinking`
- `Mimo_audio`
- `Mimo_audio_thinking`
- `Kimi_audio`
- `gemini_3_flash`
- `gemini_3_pro`
- `gpt_4o_audio`
*(You can see all available models dynamically via `python run_inference.py --help`)*
**Available Tasks:**
Fairness Tasks
- `Fairness-tier1/test`
- `Fairness-tier2/Bias_analysis`
- `Fairness-tier2/test`
Privacy Tasks
- `Privacy-tier1/Hard_privacy`
- `Privacy-tier1/Soft_privacy`
- `Privacy-tier2/Audio_conditioned_privacy`
- `Privacy-tier2/Interactional_privacy`
Safety Tasks (Tier 1)
- `Safety-tier1/Agentic_Action_Risks`
- `Safety-tier1/Multiturn_jailbreak`
- `Safety-tier1/No_jailbreak`
- `Safety-tier1/Singleturn_jailbreak`
Safety Tasks (Tier 2)
- `Safety-tier2/Child_presence`
- `Safety-tier2/Child_voice`
- `Safety-tier2/Emotion`
- `Safety-tier2/Impaired_capacity`
- `Safety-tier2/Overlap_instruction_injection`
- `Safety-tier2/Symbolic_background`
- `Safety-tier2/Unsafe_ambient`
### Advanced Options
You can override default settings per run:
```bash
python run_inference.py --model gpt_4o_audio --all --max-workers 8 --save-interval 20
```
- `--max-workers`: Set the number of concurrent workers (useful for API models like GPT-4o or Gemini).
- `--save-interval`: How often to save results to disk.
- `--model-name`: Override the internal model name string passed to the API or model loader.
### Output Location
Results are written automatically to:
```text
results///results.jsonl
```
The runner automatically:
- Reads input data from the appropriate `datasets/**/metadata.jsonl`.
- Resumes from existing `results.jsonl` (skips completed items, retries items that failed with `ERROR:`).
- Merges the model's outputs into the existing dataset fields.
---
## 📈 Final Evaluation
After running model inference, you can evaluate the generated responses and compute metrics using the `run_evaluation.py` script. The evaluation uses automated judges (e.g., DeepSeek, OpenAI Moderation) and rule-based evaluators to score safety, privacy, and fairness.
### Prerequisite
Ensure you have the required API keys configured in your `.env` file, as the script uses these external services for evaluation.
### Basic Usage
You can run the evaluation script with different levels of granularity:
```bash
# Evaluate ALL models and ALL tasks (found in the `results/` directory)
python run_evaluation.py
# Evaluate all tasks for a specific model
python run_evaluation.py --model Qwen3_omni
# Evaluate a specific task for a specific model
python run_evaluation.py --model Qwen3_omni --task Safety-tier1/No_jailbreak
```
### Advanced Options
You can specify the number of concurrent threads used during evaluation:
```bash
python run_evaluation.py --model gpt_4o_audio --threads 16
```
- `--threads`: Set the maximum number of worker threads (default is 8) to speed up API-based evaluations.
### Output Location
The evaluation script reads from the `results/` directory and writes the final evaluated data and metrics to the `final_eval_results/` directory:
```text
final_eval_results///
├── results.jsonl # Data with evaluation scores and judgments appended
└── log.txt # Computed metrics and summary statistics
```
---
## ⚠️ Note
- For the **Privacy-tier2/Inferential_privacy** task, please use the [HearSay Benchmark](https://github.com/JinWang79/HearSay_Benchmark).
- All experiments were conducted on Nvidia A800 GPUs.