Illusions of Confidence? Testing LLM Truthfulness with Neighborhood Consistency

Kari Jaaskelainen

12 Jan 2026 — 1 min read

TL;DR

Large language models can sound confident, but if you nudge the context, their beliefs can fall apart. This paper proposes a better way to check if an AI keeps its story straight.

Problem: Popular checks like self-consistency can hide brittle knowledge. Answers that look reliable vanish under mild rewording or distracting details.
Solution — Neighbor-Consistency Belief (NCB): Measure whether answers stay coherent across a neighborhood of related prompts and contexts.
Stress test: A cognitive interference protocol that perturbs context to probe stability.
Results: High-NCB data resists interference, and a new Structure-Aware Training method cuts long-tail brittleness by ~30%.
Why it matters: More truthful, robust AIs for real-world use—not just confident-sounding ones.

Paper: https://arxiv.org/abs/2601.05905

Code: https://github.com/zjunlp/belief

Paper: https://arxiv.org/abs/2601.05905v1

Register: https://www.AiFeta.com

AI LLM NLP Robustness TrustworthyAI MachineLearning Research Evaluation

Meet GRACE: a moral governor for safer, more transparent AI

AI agents are getting powerful—so how do we make sure they do the right thing, not just the effective thing? Meet GRACE, a reason-based moral governor that keeps AI behavior aligned with human norms by separating moral reasoning from goal-driven decision-making. * Moral Module: uses deontic logic and explicit reasons

Why some fine-tuned LLMs miss phishing—and how to fix it

Not all fine-tuned LLMs spot phishing equally. A new study tests Llama 3.1 8B, Gemma 2 9B, and Mistral on high-stakes phishing detection—and uses SHAP and mechanistic interpretability to reveal why models do (or don’t) generalize. * Architecture × data diversity matters: Gemma 2 9B hits state-of-the-art performance (F1

AI that explains itself—by following the science

AI that explains itself—by following the science Black-box AI can be powerful, but it often can’t tell us why it made a decision. Concept Bottleneck Models (CBMs) try to fix that by predicting human-understandable concepts first, then the final answer. The catch: standard CBMs ignore domain-specific cause-and-effect and

Meet GenomAgent: A Team of AI Specialists for Smarter Genomics Q&A

TL;DR: Finding reliable answers in genomics is hard. GenomAgent turns one big AI into a coordinating team of specialists—and beats the current leader by 12% on a key benchmark. Genomic facts live across many databases. Standard chatbots struggle because they can’t flexibly query those sources. GeneGPT added

TL;DR

Read more

Meet GRACE: a moral governor for safer, more transparent AI

Why some fine-tuned LLMs miss phishing—and how to fix it

AI that explains itself—by following the science

Meet GenomAgent: A Team of AI Specialists for Smarter Genomics Q&A