Can We Trust LLM's Logic? Quantifying Uncertainty, Coherence, and Robustness via a Graph-Based Framework
Riccardo Revalor, Jalees Rehman, Debjit Pal
Why It Matters
What makes this one worth your time
This work addresses critical issues of reasoning flaws in LLMs, providing tools that could enhance the reliability of AI systems in practical applications.
A novel framework to quantify and improve the reasoning fidelity of large language models.
Summary
The paper introduces GRAPHEVAL, a graph-based framework for quantifying uncertainty in LLM reasoning, proposing a new metric called Graph Reasoning Coherence Score (GRCS) and a decoding strategy named Graph Self-Consistency (GSC) to enhance reasoning fidelity.
Key contributions
- Development of the Graph Reasoning Coherence Score (GRCS) for measuring reasoning fidelity.
- Introduction of Graph Self-Consistency (GSC) as a decoding strategy to improve reasoning accuracy.
- Demonstration of the 'load-bearing path' concept in reasoning topology under adversarial conditions.
Notable insights
- The introduction of a graph-based approach to quantify reasoning fidelity is a unique angle that may reveal deeper insights into LLM behavior.
- The correlation of GRCS with reasoning faithfulness suggests a potential new standard for evaluating LLM outputs.
Possible limitations
- The abstract does not provide specific experimental results or comparisons to existing methods, which limits the assessment of the proposed metrics' effectiveness.
- Potential edge cases in reasoning scenarios may not be addressed.
Abstract
arXiv:2607.08017v1 Announce Type: cross Abstract: Large-Language Models (LLMs) can be prone to flawed and unfaithful reasoning that decoding strategies like Self-Consistency (SC) fail to detect as they evaluate only final-answer agreement while ignoring the logical validity of intermediate steps. This raises three fundamental questions: How can we reliably quantify uncertainty in LLM reasoning? Can semantic, structural, and causal awareness select more faithful reasoning compared to na\"ive majority voting? and How robust is reasoning topology under adversarial conditions? To address these questions, we introduce GRAPHEVAL, a graph-based reasoning framework that re-frames uncertainty quantification (UQ) as a holistic reasoning fidelity problem. We propose a novel UQ metric, Graph Reasoning Coherence Score (GRCS), that quantifies semantic-structural consensus of the reasoning space and captures pathological mode collapse and confident hallucinations. We find that GRCS is the only metric that is consistently negatively correlated with reasoning faithfulness across both more capable and smaller models. Additionally, we introduce Graph Self-Consistency (GSC), a medoid-based decoding strategy that trades nominal accuracy for reasoning fidelity, exposing the degree to which SC is inflated by unfaithful lucky guesses in smaller models, while preserving or improving accuracy in more capable ones. Finally, through adversarial medoid ablation, we demonstrate that the GSC-selected path acts as a "load-bearing path" and forcing models away from it degrades reasoning faithfulness and, in targeted cases, causes drops in accuracy.