Steer LLM Latents for Hallucination Detection
This addresses hallucination detection for safer LLM deployment, offering a practical solution with strong generalization, though it is incremental as it builds on existing latent space methods.
The paper tackles the problem of detecting hallucinations in LLM outputs by proposing a lightweight steering vector that reshapes the latent space to better separate truthful and hallucinated content, achieving state-of-the-art performance with minimal labeled data.
Hallucinations in LLMs pose a significant concern to their safe deployment in real-world applications. Recent approaches have leveraged the latent space of LLMs for hallucination detection, but their embeddings, optimized for linguistic coherence rather than factual accuracy, often fail to clearly separate truthful and hallucinated content. To this end, we propose the Truthfulness Separator Vector (TSV), a lightweight and flexible steering vector that reshapes the LLM's representation space during inference to enhance the separation between truthful and hallucinated outputs, without altering model parameters. Our two-stage framework first trains TSV on a small set of labeled exemplars to form compact and well-separated clusters. It then augments the exemplar set with unlabeled LLM generations, employing an optimal transport-based algorithm for pseudo-labeling combined with a confidence-based filtering process. Extensive experiments demonstrate that TSV achieves state-of-the-art performance with minimal labeled data, exhibiting strong generalization across datasets and providing a practical solution for real-world LLM applications.