Bias Mitigation in Audio-Enabled LLMs through Adversarial Training and Distillation Methods
In this work, we focus on the gender attribute and show that models' perception of it can be unlearned using adversarial methods without a loss in performance on other benchmarked tasks.
ProofPaper ↗
Key points
- Audio-enabled large language models (AuLLMs) allow users to interact with them simply by talking to them, without the need for an intermediate transcription.
- This expands the model's accessibility and capabilities by encoding non-verbal cues.
- However, using non-verbal cues may lead to biased behaviour based on the attributes inferred (or assumed) from them.
- Such models are useful in situations where gender inference is unwanted.
Sources (1)
- [1]Bias Mitigation in Audio-Enabled LLMs through Adversarial Training and Distillation MethodsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 15, 12:19 PM
In this work, we focus on the gender attribute and show that models' perception of it can be unlearned using adversarial methods without a loss in performance on other benchmarked tasks.
Audio-enabled large language models (AuLLMs) allow users to interact with them simply by talking to them, without the need for an intermediate transcription.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 15, 2026Anthropic tells investors quarterly revenue up nearly 50% to $16.8bn ahead of IPO
- Oct 14, 2026Recursive Offline Distillation via Entropy Gain Sampling and Answer Mixing
- Oct 14, 2026Who's Behind the Text? Robust Model Authorship Attribution
- Oct 14, 2026Beyond Corrected Actions: Training LLM Agents on Textual Critiques
- Oct 14, 2026RuleTrek: Benchmarking Agentic Rule Reasoning in Complex Scenarios
- Oct 13, 2026Plan or Retrieve First? A Scale Study of Tool Discovery Architectures
