Directional Concentration Uncertainty: A representational approach to uncertainty quantification for generative models

TL;DR

This paper introduces Directional Concentration Uncertainty (DCU), a flexible statistical method for quantifying uncertainty in generative models by analyzing the geometric dispersion of embeddings, improving trustworthiness and robustness across tasks.

Contribution

The paper presents a novel, task-agnostic UQ framework using the von Mises-Fisher distribution to measure embedding dispersion, outperforming heuristic methods in generative model uncertainty quantification.

Findings

DCU matches or exceeds calibration of prior methods like semantic entropy.

DCU generalizes well to complex multi-modal tasks.

Framework supports integration into multi-modal and agentic systems.

Abstract

In the critical task of making generative models trustworthy and robust, methods for Uncertainty Quantification (UQ) have begun to show encouraging potential. However, many of these methods rely on rigid heuristics that fail to generalize across tasks and modalities. Here, we propose a novel framework for UQ that is highly flexible and approaches or surpasses the performance of prior heuristic methods. We introduce Directional Concentration Uncertainty (DCU), a novel statistical procedure for quantifying the concentration of embeddings based on the von Mises-Fisher (vMF) distribution. Our method captures uncertainty by measuring the geometric dispersion of multiple generated outputs from a language model using continuous embeddings of the generated outputs without any task specific heuristics. In our experiments, we show that DCU matches or exceeds calibration levels of prior works like semantic entropy (Kuhn et al., 2023) and also generalizes well to more complex tasks in multi-modal domains. We present a framework for the wider potential of DCU and its implications for integration into UQ for multi-modal and agentic frameworks.