Complementing Self-Consistency with Cross-Model Disagreement for Uncertainty Quantification

2026-04-18

AI Summary

This work addresses the challenge that large language models often generate erroneous responses with high confidence, undermining uncertainty estimation based solely on self-consistency. To overcome this limitation, the authors propose a black-box framework for estimating total uncertainty without requiring access to internal model parameters or output probability distributions. The approach leverages a small ensemble of similarly sized models, combining intra-model self-consistency (aleatoric uncertainty) with inter-model semantic disagreement (epistemic uncertainty). Experiments across five instruction-tuned models (7–9B parameters) and ten long-text tasks demonstrate that the method substantially outperforms baselines relying only on aleatoric uncertainty, particularly in detecting high-confidence yet incorrect predictions. This leads to improved calibration and enhanced selective abstention capabilities.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Uncertainty Representations
• Multiagent Systems: Multiagent Systems under Uncertainty

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Large language models (LLMs) often produce confident yet incorrect responses, and uncertainty quantification is one potential solution to more robust usage. Recent works routinely rely on self-consistency to estimate aleatoric uncertainty (AU), yet this proxy collapses when models are overconfident and produce the same incorrect answer across samples. We analyze this regime and show that cross-model semantic disagreement is higher on incorrect answers precisely when AU is low. Motivated by this, we introduce an epistemic uncertainty (EU) term that operates in the black-box access setting: EU uses only generated text from a small, scale-matched ensemble and is computed as the gap between inter-model and intra-model sequence-semantic similarity. We then define total uncertainty (TU) as the sum of AU and EU. In a comprehensive study across five 7-9B instruction-tuned models and ten long-form tasks, TU improves ranking calibration and selective abstention relative to AU, and EU reliably flags confident failures where AU is low. We further characterize when EU is most useful via agreement and complementarity diagnostics.

Problem

• uncertainty quantification
• large language models
• self-consistency
• aleatoric uncertainty
• epistemic uncertainty

Innovation

• epistemic uncertainty
• cross-model disagreement
• self-consistency
• uncertainty quantification
• black-box ensemble