Deliberative Reasoning Network: An Uncertainty-Driven Paradigm for Belief-Tracked Inference with Pretrained Language Models

2025-08-06

AI Summary

Large language models (LLMs) frequently succumb to cognitive biases when semantic heuristics conflict with decisive evidence, leading to logical reasoning failures. To address this, we propose “Uncertainty Minimization” — a novel paradigm replacing conventional probability maximization — that enables interpretable and verifiable System 2–style reasoning via explicit belief-state tracking and iterative evidence synthesis. Our method comprises four core components: an evidence-integration–based belief tracking network, an uncertainty quantification module, a lightweight verification component, and a dedicated discriminative model, all fully compatible with mainstream generative LLMs. The framework supports zero-shot transfer and achieves a 15.2% absolute gain over baselines on the adversarial reasoning benchmark LCR-1000. Integrated with Mistral-7B, it raises accuracy on challenging reasoning tasks from 20% to 80%, and improves zero-shot performance on TruthfulQA by 23.6%.

Technology Category

• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Machine Learning: Calibration & Uncertainty Quantification
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Large language models often fail at logical reasoning when semantic heuristics conflict with decisive evidence - a phenomenon we term cognitive traps. To address this fundamental limitation, we introduce the Deliberative Reasoning Network (DRN), a novel paradigm that reframes logical reasoning from probability maximization to uncertainty minimization. Instead of asking "Which answer is most likely?", DRN asks "Which hypothesis has the most internally consistent evidence?". DRN achieves intrinsic interpretability by explicitly tracking belief states and quantifying epistemic uncertainty for competing hypotheses through an iterative evidence synthesis process. We validate our approach through two complementary architectures - a bespoke discriminative model that embodies the core uncertainty minimization principle, and a lightweight verification module that enhances existing generative LLMs. Evaluated on LCR-1000, our new adversarial reasoning benchmark designed to expose cognitive traps, the bespoke DRN achieves up to 15.2% improvement over standard baselines. When integrated as a parameter-efficient verifier with Mistral-7B, our hybrid system boosts accuracy from 20% to 80% on the most challenging problems. Critically, DRN demonstrates strong zero-shot generalization, improving TruthfulQA performance by 23.6% without additional training, indicating that uncertainty-driven deliberation learns transferable reasoning principles. We position DRN as a foundational, verifiable System 2 reasoning component for building more trustworthy AI systems.

Problem

• Addresses logical reasoning failures in large language models
• Minimizes uncertainty instead of maximizing probability in reasoning
• Enhances interpretability by tracking belief states and uncertainty

Innovation

• Uncertainty minimization replaces probability maximization
• Tracks belief states for interpretable reasoning
• Lightweight verifier enhances existing generative LLMs