Artificial Intelligence

2601 Submissions

[5] ai.viXra.org:2601.0117 [pdf] submitted on 2026-01-29 18:48:52

Geodesics of Meaning: Modeling Semantic Curvature in Transformer-Based Language Models via General Relativity

Authors: Travis Shane Taylor
Comments: 37 Pages.

We present a general relativistic framework for modeling transformer-based language models (LLMs) as nonlinear dynamical systems evolving on curved semantic manifolds. Standard transformer architectures are shown to approximate a flat Minkowski spacetime, where attention mechanisms define a local semantic metric tensor. We extend this formulation by introducing curved metrics—specifically the Schwarzschild and Friedmann—Lemaître—Robertson—Walker (FLRW) solutions—to model context-sensitive meaning, narrative curvature, and long-range semantic dependencies. A stress-energy tensor encodes topical mass, tonal flow, and tension, driving semantic curvature via Einstein’s field equations. We validate this framework using both simplified language simulations and full narrative data, showing that Ricci curvature serves as a physically interpretable measure of coherence, complexity, and twist. This work bridges differential geometry, nonlinear systems, and AI interpretability, offering a new paradigm for analyzing and guiding large language model behavior.
Category: Artificial Intelligence

[4] ai.viXra.org:2601.0076 [pdf] submitted on 2026-01-18 22:45:21

The Federation of Selves: A Framework for Human Agency and State-Dependent Alignment in High-Stakes AI Environments

Authors: Natasha Zink
Comments: 3 Pages. (Note by ai.viXra.org Admin: Please cite listed scientific references; the scientific references are not listed in a complete/standard manner such as APS style)

As Artificial Intelligence scales toward autonomous decision-making, the problem of "alignment" remains the primary obstacle to safe implementation. We propose a model of "Human-AI Husbandry," moving away from autonomous agentstoward "Motile Utilities." By treating human intent as a dynamic, state-dependentsystem via Fourier-Laplace transforms, we introduce the Federation of Selves—anarchitecture that magnifies human agency while preserving the Lead's sovereignty through biometric and psychological interlocks.
Category: Artificial Intelligence

[3] ai.viXra.org:2601.0055 [pdf] submitted on 2026-01-14 04:26:00

The One-to-One Resolution: Deterministic $O(N^2)$ Synthesis of the Traveling Salesman Problem via the Unseen Syntax

Authors: Chaiya Tantisukarom
Comments: 4 Pages.

The Traveling Salesman Problem (TSP) has long been the hallmark of $NP$-hard complexity. This paper presents a definitive shift in the problem's resolution by moving from a universal search-based paradigm to a "One-to-One" synthesis approach. By utilizing the "Unseen Syntax" algorithm—a deterministic $O(N^2)$ procedure applied to a specific, fixed-start matrix—we demonstrate that the complexity of a localized reality is polynomial. We further argue that the perceived $NP$-hardness is a result of a dimensional mismatch between the specific problem instance and the search for a universal solution.
Category: Artificial Intelligence

[2] ai.viXra.org:2601.0032 [pdf] submitted on 2026-01-11 15:40:44

The Verification Trap: Epistemic Decay and the Rise of Fabricated Truth in the Post-GenAI Era

Authors: Chaiya Tantisukarom
Comments: 8 Pages. (Note by ai.viXra.org Admin: Author name should be listed in the article after the article title))

This paper formalizes the "Visual Warning Map" theorem, detailing the intersection of AI Signal Value and Grounded Resource Costs. We argue that without proactive provenance anchoring, the information ecosystem will reach an "Epistemic Crossover" point between Year 5 and Year 7. Post-Year 7, "Fabricated Truth" becomes the dominant global heuristic, leading to a catastrophic "Verification Debt" that threatens the stability of human-digital governance.
Category: Artificial Intelligence

[1] ai.viXra.org:2601.0006 [pdf] replaced on 2026-01-03 17:43:27

Lossless Vessel: Unconditional Analytic Closure for the Riemann Hypothesis

Authors: Julio C. Luna
Comments: 95 Pages. Updated to [include] a formal Abstract, Keywords, and detailed Table of Contents. Added a comprehensive Bibliography and moved computational reproducibility logs to Appendix Y to ensure deductive clarity.

We present the Lossless Vessel framework, an operator-theoretic approach to establishing an unconditional analytic closure for the Riemann Hypothesis (RH). The argument is organized into three firewalled domains: (A) a fully deductive proof, (B) conceptual interpretation, and (C) computational reproducibility artifacts. In Domain A we reduce RH to an Execution Bound EB(ε) and prove EB(ε) uniformly for all ε>0 via a defect-to-Carleson control mechanism for the arithmetic boundary field, concluding that every non-trivial zero ρ of ζ(s) satisfies Re(ρ)=1/2. The proof draws on standard tools from analytic number theory and harmonic analysis, including Hardy-space embeddings and Carleson measure estimates (see, e.g., [1—5]). Computational material is included only for auditability and is not used as a logical premise.
Category: Artificial Intelligence