Non-Linear Inference Time Intervention: Improving LLM Truthfulness
This work addresses truthfulness in LLMs for users relying on accurate AI outputs, representing an incremental improvement over existing ITI methods.
The authors tackled the problem of improving LLM truthfulness by developing Non-Linear Inference Time Intervention (NL-ITI), a method that enhances performance on benchmarks like TruthfulQA, achieving over 16% relative improvement in accuracy compared to baseline ITI and 10% over a recent method.
In this work, we explore LLM's internal representation space to identify attention heads that contain the most truthful and accurate information. We further developed the Inference Time Intervention (ITI) framework, which lets bias LLM without the need for fine-tuning. The improvement manifests in introducing a non-linear multi-token probing and multi-token intervention: Non-Linear ITI (NL-ITI), which significantly enhances performance on evaluation benchmarks. NL-ITI is tested on diverse multiple-choice datasets, including TruthfulQA, on which we report over 16% relative MC1 (accuracy of model pointing to the correct answer) improvement with respect to the baseline ITI results. Moreover, we achieved a 10% relative improvement over the recently released Truth Forest (TrFf) method that also focused on ITI improvement.