large language models

The Evidence You Do Not Give the AI: Holdout Asymmetry and Epistemic Self-Correction Under Model Mismatch

This work introduces holdout asymmetry: the epistemic asymmetry created when a human evaluator can reserve evidence outside a language model's conversational context and later reveal it after the model has committed to an explanatory position. It develops the associated concepts of prediction vacancy and degenerating a...

Can an LLM Agree with Itself? Repeatability and Reproducibility in Grant Screening

Large language models are increasingly being considered for grant proposal screening, yet their reliability has received less attention than their agreement with human reviewers. This case study evaluates the repeatability and reproducibility of direct textual interpretation by DeepSeek-V4-Flash-0731, run in reasoning...

The Free-Energy Architecture of Discovery-Capable Intelligence

Large language models optimise fluently inside a fixed representational space but do not reliably produce genuine discovery; the field has treated this as a tuning problem. We argue it is an architectural one, and we derive the required architecture from first principles rather than inferring it from current systems. T...

AgroYachay: An open-source IoT and large-language-model platform supporting agronomic and economic decision-making for Andean smallholders

Metadata release accompanying the SoftwareX submission. Changed Author affiliations synced with the manuscript, as confirmed by the authors: Fred Torres-Cruz — Doctoral Program in Sustainable Agriculture, UNALM Eduardo Fuentes-Navarro — Faculty of Animal Science, UNALM UNAP department spelled out; corresponding author...

Reality Drift Tools: Evaluation, Diagnostics, and Governance for AI Systems

This collection presents three practical tools developed as part of the Semantic Fidelity Project for evaluating Reality Drift in artificial intelligence systems, including large language models, retrieval systems, and multi-step agent workflows.Reality Drift occurs when a system continues to operate coherently while g...

Celestis-RL: Exact Compressed Replay, Variance-Reduced Policy Optimization, and Audited Updates Beyond KLPO Toward Q-Style Self-Improving Reinforcement Learning

Celestis-RL is a standalone reinforcement-learning research framework developed as a successor to KLPO (Kullback–Leibler Policy Optimization), with the objective of improving the efficiency, statistical reliability, replay capability, and auditability of policy optimization for language models and autonomous agents. It...

NLP-Digitox: An NLP and LLM-Driven Mobile Application for Managing Smartphone Overuse Among Students

NLP-Digitox is a privacy-oriented mobile application designed to help college students manage excessive smartphone use through behavioral monitoring and personalized intervention. The system combines on-device smartphone usage tracking with Natural Language Processing (NLP), sentiment analysis, and Large Language Model...

SABRINAL EL-RAKHAWI DOCUMENT: CONSCIOUSNESS, ARTIFICIAL INTELLIGENCE, AND LAW Foundations: Biology, Mind, and Legal Systems

This foundational document explores the intersection of consciousness, artificial intelligence, and law through three comprehensive books. Book One traces biological evolution from primordial replicators to human consciousness, examining genetic engineering and legal personhood. Book Two investigates artificial intelli...

Clinical quality of large language model answers to questions about pancreatic cancer

Non-Equilibrium Geometric Field Theory of In-Silico Intelligence: Discrete Ricci Flow, Holographic Self-Encoding, and O(N) Autopoietic Cognition on Hypergraph Manifolds

We propose and establish a framework for post-transformer (post-llm) architecture within a new paradigm. Transformer architecture and large language models are libraries with probabilistic indices. The brain never constituted either intelligence, nor consciousness - not even processing. Brain processing is a category e...

SISTEMA VARIACIONAL BIOFÍSICO PARA TECNOLOGIAS AVANÇADAS DE BATERIAS

SVBAC-DE: Modelo Multifísico e de Sistema para Baterias de Alto Desempenho — Uma Proposta de Modelagem Quântica, Predição e Controle Adaptativo Este estudo propõe uma arquitetura conceitual e computacional para a modelagem de baterias de alto desempenho baseada no Sistema Variacional Biofísico Acoplado da Dinâmica Elet...

Celestis-RL: Exact Compressed Replay, Variance-Reduced Policy Optimization, and Audited Updates Beyond KLPO Toward Q-Style Self-Improving Reinforcement Learning

Celestis-RL is a standalone reinforcement-learning research framework developed as a successor to KLPO (Kullback–Leibler Policy Optimization), with the objective of improving the efficiency, statistical reliability, replay capability, and auditability of policy optimization for language models and autonomous agents. It...