Search papers, labs, and topics across Lattice.
This paper introduces Prediction of Prediction (PoP), a novel mechanism that leverages inter-layer activation fusion to detect factual inaccuracies in large language models (LLMs) during a single forward pass. By analyzing hidden-state transition dynamics, PoP effectively identifies overconfident erroneous outputs without the need for additional decoding, achieving an AUROC of 75.5% on the TruthfulQA benchmark. This approach significantly reduces memory and latency overhead compared to traditional multi-sample verification methods, making it a practical solution for enhancing the reliability of LLMs in critical applications.
Internal activation dynamics can reveal factual errors in LLM outputs without the costly overhead of multiple decoding passes.
Autoregressive large language models (LLMs) routinely generate factually incorrect outputs with high decoding confidence, limiting their deployment in high-stakes workflows. Existing output-stage uncertainty metrics can fail when models are overconfident on false assertions, while multi-sample verification pipelines introduce substantial memory and latency overhead. This work evaluates whether internal hidden-state transition dynamics during generation can signal factual errors without auxiliary decoding calls. We introduce Prediction of Prediction (PoP), a mechanism that captures layer-transition uncertainty by fusing intermediate hidden representations across depth during a single forward pass. Evaluated on the TruthfulQA benchmark using autoregressive transformer backbones, PoP achieves an area under the receiver operating characteristic curve (AUROC) of 75.5% for factual-correctness classification. The mechanism operates within the base forward pass, adding less than 1.2% runtime latency and requiring zero additional generation passes. The numerical results are reported from the author-verified experimental implementation and are bounded by the evaluation scope described below.
