A Causal Explainable Guardrails for Large Language Models
This work addresses the issue of bias in LLMs for users seeking safe and reliable AI systems, but it appears incremental as it builds on existing steering methods by adding causal and adversarial components.
The paper tackles the problem of undesirable attributes or biases in Large Language Models (LLMs) by proposing LLMGuardrail, a framework that uses causal analysis and adversarial learning to obtain unbiased steering representations, resulting in effective steering toward desired attributes while mitigating biases.
Large Language Models (LLMs) have shown impressive performance in natural language tasks, but their outputs can exhibit undesirable attributes or biases. Existing methods for steering LLMs toward desired attributes often assume unbiased representations and rely solely on steering prompts. However, the representations learned from pre-training can introduce semantic biases that influence the steering process, leading to suboptimal results. We propose LLMGuardrail, a novel framework that incorporates causal analysis and adversarial learning to obtain unbiased steering representations in LLMs. LLMGuardrail systematically identifies and blocks the confounding effects of biases, enabling the extraction of unbiased steering representations. Additionally, it includes an explainable component that provides insights into the alignment between the generated output and the desired direction. Experiments demonstrate LLMGuardrail's effectiveness in steering LLMs toward desired attributes while mitigating biases. Our work contributes to the development of safe and reliable LLMs that align with desired attributes.