🎧 Awesome Trustworthy Audio-LLMs

A curated and continuously evolving collection of research papers, benchmarks, datasets, and open resources on Trustworthy Audio Large Language Models, covering the full spectrum of safety, robustness, and reliability in audio large models. The collection also includes representative works on audio large models beyond the safety domain.

🌍Introduction

With the rapid developmenst of Audio Large Language Models (Audio-LLMs), ensuring their trustworthiness and safety has become an essential research frontier.

This repository(TALLM) serves as a comprehensive and community-driven hub for tracking progress in the field of trustworthy audio intelligence.

It highlights research on:

• 🛡️ Safety
• ⚖️ Fairness
• 🔮 Hallucination
• 🔐 Privacy
• ⚙️ Robustness
• 💡 Interpretability
• 💻 Security
• …

Together, these works aim toward a future where Audio-LLMs are not only capable of understanding voices — but also worthy of being trusted.

📌 Table of Content

• 🧭 Research Collections
• 🎒Good Papers
• 🗞️ Recent News
• 🤝 How to Contribute
• 💬 LLM Safety Discussion
• 🌟 Rising Stars
• 🙏Acknowledgement

🧭 Research Collections

• Paper
  • A. Safety
    • 📖A1. General
    • ✏️A2. Jailbreak
    • 🏫A3. Alignment
    • 📚A4. Deepfake
    • 🎒A5. Prompt Injection
    • 🧑‍🏫A6. Defense
  • B. Security
    • 📖B1. General
    • ✏️B2. Adversarial Examples
    • 🏫B3. Attack
    • 📚B4. Poison & Backdoor
  • C. Privacy
    • 📖C1. General
  • D. Interpretability
    • 📖D1. General
  • E. Fairness
    • 📖E1. General
  • F Hallucination
    • 📖F1. General
  • G Robustness
    • 📖G1.General

In addition to the above-mentioned ones:

If you want to learn more about Audio Large Language Models, you can take a look at the following.

• 🚀 Survey
• 🚀 Technical Report for Audio Large Language Models
• 🚀 Toolkit
• 🚀 Benchmark
• 🚀 Capability
• 🚀 TTS
• 🚀 General Papers in Audio
• 🚀 Traditional Models

🎒 Good Papers

• 🏆 ChronosAudio: A Comprehensive Long-Audio Benchmark for Evaluating Audio-Large Language Models
• 🏆 ViSAudio: End-to-End Video-Driven Binaural Spatial Audio Generation
• 🏆 Attacker’s Noise Can Manipulate Your Audio-based LLM in the Real World
• 🏆 Speech-Audio Compositional Attacks on Multimodal LLMs and Their Mitigation with SALMONN-Guard
• 🏆 AHa-Bench: Benchmarking Audio Hallucinations in Large Audio-Language Models
• 🏆 Hidden in the Noise: Unveiling Backdoors in Audio LLMs Alignment through Latent Acoustic Pattern Triggers

🗞️ Recent News

• [2025.11.12] 🐣TALLM is released!!!
• [2025.11.25] ⛰️Version1 is open to everyone!!!
• [2025.11.26] 🍠TALLM has posted on xhs!!!
• [2025.12.01] 📈General capabilities have been collected!!!
• [2025.12.08] 🌴TTS and Other in Audio have been collected!!!
• [2025.12.11] 😈TALLM has 100+ Stars!!!
• [2025.12.29] 🆕2025.12 papers are collected!!!
• [2026.03.18] 🆕2026.01 - 2026.03 papers are collected!!!
• [2026.04.14] 🆕2026.04 papers are collected!!!
• [2026.04.25] 🐶Version2 is open to everyone!!!
• [2026.05.21] 📑Survey Paper is uploaded!!!

🤝 How to Contribute

We welcome contributions from researchers and practitioners! If you’d like to submit a piece of writing, please write an email to kaiwenluo74@gmail.com or write your paper link in issues part. Then we can add the article in.

💬 LLM Safety Discussion

🌟 Rising Stars

🙏 Acknowledgement

• Organizers: Kevin Luo, Zhenhong Zhou, Liang Lin, Yuting Ruan, Yuanhe Zhang, Tianyu Shao
• Thank you to all those who have followed this project.