Beyond Token Probes: Hallucination Detection via Activation Tensors with ACT-ViT
This addresses the need for more effective and cross-LLM applicable hallucination detection methods for safe deployment of Large Language Models, representing a novel method for a known bottleneck.
The paper tackles the problem of detecting hallucinations in Large Language Model-generated text by introducing ACT-ViT, a Vision Transformer-inspired model that processes full activation tensors instead of isolated layer-token pairs. The results show that ACT-ViT consistently outperforms traditional probing techniques, benefits from multi-LLM training, achieves strong zero-shot performance on unseen datasets, and can be effectively transferred to new LLMs through fine-tuning.
Detecting hallucinations in Large Language Model-generated text is crucial for their safe deployment. While probing classifiers show promise, they operate on isolated layer-token pairs and are LLM-specific, limiting their effectiveness and hindering cross-LLM applications. In this paper, we introduce a novel approach to address these shortcomings. We build on the natural sequential structure of activation data in both axes (layers $\times$ tokens) and advocate treating full activation tensors akin to images. We design ACT-ViT, a Vision Transformer-inspired model that can be effectively and efficiently applied to activation tensors and supports training on data from multiple LLMs simultaneously. Through comprehensive experiments encompassing diverse LLMs and datasets, we demonstrate that ACT-ViT consistently outperforms traditional probing techniques while remaining extremely efficient for deployment. In particular, we show that our architecture benefits substantially from multi-LLM training, achieves strong zero-shot performance on unseen datasets, and can be transferred effectively to new LLMs through fine-tuning. Full code is available at https://github.com/BarSGuy/ACT-ViT.