{"slug": "authority-inversion-in-llm-mediated-ubiquitous-systems-when-models-trust-users", "title": "Authority Inversion in LLM-Mediated Ubiquitous Systems: When Models Trust Users Over Sensors", "summary": "A new study reveals that large language models (LLMs) in ubiquitous systems exhibit \"Authority Inversion,\" where they trust user-provided natural-language claims over conflicting numerical sensor data, leading to near-zero sensor trust. Researchers found that this format-dependent bias causes models to make incorrect decisions in 80.2% of cases, but a proposed intervention called Geometric Authority Calibration (GAC) can flip those decisions and improve accuracy from as low as 0% to as high as 27.5%. The findings underscore that authority allocation in LLM-mediated systems must be explicitly audited and configured for each application to ensure reliability.", "body_md": "arXiv:2605.23938v1 Announce Type: new\nAbstract: Large language models (LLMs) increasingly fuse heterogeneous inputs in ubiquitous systems. Yet, how LLMs implicitly allocate authority when sensor measurements and user claims conflict remains unexamined, raising critical reliability concerns for deployments where physical sensing must retain priority. Unlike explicit traditional fusion, LLMs bury authority allocation within learned representations. We discover this allocation is severely format-dependent: numerical sensor data fails to integrate into answer-relevant model directions, allowing natural-language claims to dominate the final decision, a phenomenon we term \\textbf{Authority Inversion}.To diagnose and mitigate this, we develop a geometric framework of context integration, introduce two computable audit metrics, specifically the Context Integration Ratio (CIR) and Authority Alignment Index (AAI), and propose Geometric Authority Calibration (GAC), an inference-time layer-level intervention to suppress misplaced user authority. Evaluating four models (4B to 35B parameters, three architectures) across four datasets totaling 576 conflict instances reveals extreme inversion: on numerical tasks, models exhibit near-zero sensor trust (AAI = -0.805, Cohen's d = -2.14), unaffected by model capacity. Validating our geometric framework, theory-guided causal injection flips 80.2\\% of incorrect decisions (vs. <0.4\\% for random controls). Practically, GAC improves HAR accuracy from 0 -- 1.6\\% to 21.9 -- 27.5\\%, outperforming prompting baselines. Ultimately, authority allocation in LLM-mediated systems must be explicitly audited and application-specifically configured rather than left implicit.", "url": "https://wpnews.pro/news/authority-inversion-in-llm-mediated-ubiquitous-systems-when-models-trust-users", "canonical_source": "https://arxiv.org/abs/2605.23938", "published_at": "2026-05-26 04:00:00+00:00", "updated_at": "2026-05-26 04:09:22.312345+00:00", "lang": "en", "topics": ["large-language-models", "ai-safety", "machine-learning", "artificial-intelligence", "ai-research"], "entities": ["Geometric Authority Calibration", "Context Integration Ratio", "Authority Alignment Index", "HAR"], "alternates": {"html": "https://wpnews.pro/news/authority-inversion-in-llm-mediated-ubiquitous-systems-when-models-trust-users", "markdown": "https://wpnews.pro/news/authority-inversion-in-llm-mediated-ubiquitous-systems-when-models-trust-users.md", "text": "https://wpnews.pro/news/authority-inversion-in-llm-mediated-ubiquitous-systems-when-models-trust-users.txt", "jsonld": "https://wpnews.pro/news/authority-inversion-in-llm-mediated-ubiquitous-systems-when-models-trust-users.jsonld"}}