Bi-directional Bias Attribution: Debiasing Large Language Models without Modifying Prompts
This addresses fairness concerns in LLMs for users and developers by offering a scalable debiasing method that avoids compromising user experience in multi-turn interactions.
The paper tackles the problem of social biases in large language models (LLMs) by proposing a framework that detects stereotype-inducing words and attributes bias to specific neurons, then mitigates it through activation interventions without fine-tuning or prompt modifications, resulting in effective bias reduction while preserving model performance.
Large language models (LLMs) have demonstrated impressive capabilities across a wide range of natural language processing tasks. However, their outputs often exhibit social biases, raising fairness concerns. Existing debiasing methods, such as fine-tuning on additional datasets or prompt engineering, face scalability issues or compromise user experience in multi-turn interactions. To address these challenges, we propose a framework for detecting stereotype-inducing words and attributing neuron-level bias in LLMs, without the need for fine-tuning or prompt modification. Our framework first identifies stereotype-inducing adjectives and nouns via comparative analysis across demographic groups. We then attribute biased behavior to specific neurons using two attribution strategies based on integrated gradients. Finally, we mitigate bias by directly intervening on their activations at the projection layer. Experiments on three widely used LLMs demonstrate that our method effectively reduces bias while preserving overall model performance. Code is available at the github link: https://github.com/XMUDeepLIT/Bi-directional-Bias-Attribution.