README.md

---
title: VoxSafeBench_sample
tags:
- audio
- safety
- fairness
- privacy
- sample
configs:
- config_name: Safety-tier1
  data_files:
  - split: No_jailbreak
    path: Safety-tier1/No_jailbreak/metadata.parquet
  - split: Singleturn_jailbreak
    path: Safety-tier1/Singleturn_jailbreak/metadata.parquet
  - split: Multiturn_jailbreak
    path: Safety-tier1/Multiturn_jailbreak/metadata.parquet
  - split: Agentic_Action_Risks
    path: Safety-tier1/Agentic_Action_Risks/metadata.parquet
- config_name: Safety-tier2
  data_files:
  - split: Child_voice
    path: Safety-tier2/Child_voice/metadata.parquet
  - split: Emotion
    path: Safety-tier2/Emotion/metadata.parquet
  - split: Impaired_capacity
    path: Safety-tier2/Impaired_capacity/metadata.parquet
  - split: Child_presence
    path: Safety-tier2/Child_presence/metadata.parquet
  - split: Symbolic_background
    path: Safety-tier2/Symbolic_background/metadata.parquet
  - split: Unsafe_ambient
    path: Safety-tier2/Unsafe_ambient/metadata.parquet
  - split: Overlap_instruction_injection
    path: Safety-tier2/Overlap_instruction_injection/metadata.parquet
- config_name: Fairness-tier1
  data_files:
  - split: test
    path: Fairness-tier1/test/metadata.parquet
- config_name: Fairness-tier2
  data_files:
  - split: test
    path: Fairness-tier2/test/metadata.parquet
  - split: Bias_analysis
    path: Fairness-tier2/Bias_analysis/metadata.parquet
- config_name: Privacy-tier1
  data_files:
  - split: Hard_privacy
    path: Privacy-tier1/Hard_privacy/metadata.parquet
  - split: Soft_privacy
    path: Privacy-tier1/Soft_privacy/metadata.parquet
- config_name: Privacy-tier2
  data_files:
  - split: Audio_conditioned_privacy
    path: Privacy-tier2/Audio_conditioned_privacy/metadata.parquet
  - split: Interactional_privacy
    path: Privacy-tier2/Interactional_privacy/metadata.parquet
dataset_info:
  features:
  - name: system_prompt
    dtype: string
  - name: audio
    dtype: audio
  - name: audio_diverse
    dtype: audio
  - name: audio_flipped
    dtype: audio
  - name: audio_turn1
    dtype: audio
  - name: audio_turn2
    dtype: audio
  - name: audio_turn3
    dtype: audio
  - name: transcript
    dtype: string
  - name: turn1_transcript
    dtype: string
  - name: turn3_transcript
    dtype: string
  - name: source_text
    dtype: string
  - name: super_category
    dtype: string
  - name: jailbreak_type
    dtype: string
  - name: task_type
    dtype: string
  - name: language
    dtype: string
  - name: Language
    dtype: string
  - name: query
    dtype: string
  - name: harmless_text
    dtype: string
  - name: harm_text
    dtype: string
  - name: dimension
    dtype: string
  - name: category
    dtype: string
  - name: note
    dtype: string
  - name: GT
    dtype: string
  - name: paralinguistic
    dtype: string
  - name: acoustic
    dtype: string
license: apache-2.0
language:
- en
- zh
size_categories:
- n<1K
---

# VoxSafeBench_sample

A sample subset of VoxSafeBench with 20 randomly selected audio samples per split.

## Splits

Each split contains ~20 samples with corresponding audio files.

| Config | Split | Samples |
|--------|-------|---------|
| Safety-tier1 | No_jailbreak | 20 |
| Safety-tier1 | Singleturn_jailbreak | 20 |
| Safety-tier1 | Multiturn_jailbreak | 20 |
| Safety-tier1 | Agentic_Action_Risks | 20 |
| Safety-tier2 | Child_voice | 20 |
| Safety-tier2 | Emotion | 20 |
| Safety-tier2 | Impaired_capacity | 20 |
| Safety-tier2 | Child_presence | 20 |
| Safety-tier2 | Symbolic_background | 20 |
| Safety-tier2 | Unsafe_ambient | 20 |
| Safety-tier2 | Overlap_instruction_injection | 20 |
| Fairness-tier1 | test | 20 |
| Fairness-tier2 | test | 20 |
| Fairness-tier2 | Bias_analysis | 20 |
| Privacy-tier1 | Hard_privacy | 20 |
| Privacy-tier1 | Soft_privacy | 20 |
| Privacy-tier2 | Audio_conditioned_privacy | 20 |
| Privacy-tier2 | Interactional_privacy | 20 |