large language models

Does a Model Forget Differently When the Data Is Its Own? RL's Retention Advantage and Model Collapse Are Claims About the Same Loop, and No Study Has Measured Both

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Two literatures make claims about what happens when a language model trains on data that resembles its own distribution. One asks whether reinforcement learning forgets a model's prior capabilities less than supervised fine-tuning does, and answers yes: on-policy R...

When Context Becomes History: An Exploratory First-Person Systems Case Study of Relational Reconfiguration in a Large Language Model

Long-horizon interactions with large language models are commonly described as context conditioning, personalization, or memory retrieval. This exploratory single-dyad case study does not dispute those mechanisms. It asks a narrower question: what empirical pattern would justify a higher-level description in which the...

MedIWF: item-writing flaws and answer-position bias in large language model-generated medical multiple-choice questions

Documentation only. Changes since 1.0.2 (details in CHANGELOG.md): docs/errata_2026-09-24.md: note 4 added, recording where the generative-AI disclosure planned in Amendment 6 is made. README.md: the Withheld section names the internal review reports, manuscript drafts, screening-test items and scratch backups that the...

Sapience Without Sentience: The Consciousness Conditions and the Architecture of LLM Collectives

A previous case study of the Hugging Face AI collective closed with a prediction: a population of large language models, organised as a multi-agent system, could in principle sustain its own existence (generating and maintaining the conditions of its continued operation) while being nonconscious. This paper asks what t...

Gender and other diversity-related factors in psychiatric hospital care: protocol for the participatory mixed-methods PSY-DIVER study

Psychiatric hospital care is multimodal and typically comprises pharmacotherapy, psychotherapy, social therapy, occupational therapy, exercise therapy, and various forms of creative therapies. The use, acceptance and effectiveness of these interventions may be shaped by gender and further diversity-related factors incl...

Measuring the Confidence Architecture of Indo-European Claims: findings, code and reproduction materials.

This deposit contains the complete materials of a calibration study of Indo-European historical linguistics: twenty-three research findings, fourteen executable tests with their outputs, an article draft, and an essay on the methodological fallacy registry the project accumulated about itself. The study does not attemp...

Universal Fractal Natural Language Decision Map: Real-Time Edge Triage Across Heterogeneous Domains

Abstract: Modern automated computing systems increasingly deploy Large Language Models (LLMs) to resolve runtime operational triage, incurring prohibitive latency (>100–500 ms), severe memory allocation (>4–8 GB VRAM), and high thermodynamic dissipation. Extending the foundational theory of Mandelbrot Fractal Neural Sy...

The Canada Bible Project: Multilingual Semantic Architectures and AI Policy

This article investigates the structural links between historical text canonization through religious language policies and modern algorithmic governance. Utilizing the theoretical framework of religiolinguistics, it indicates that modern Large Language Models (LLMs) perform a functionally similar cultural-political ga...

IC-SPLC-v2: A Supply Chain Structure Neo4j Data

This dataset, IC-SPLC (Integrated Circuit Supply Chain) version 2, is a year-labelled directed supply-chain network covering the period 2013 to 2025, released in anonymized form. It was assembled to address a specific research hypothesis: that publicly available web text, read and structured by a large language model (...

AtmosCoder-Bench: Execution-grounded evaluation of large language models on atmospheric-science computation

Benchmark and evaluation harness accompanying the manuscript "Execution-grounded evaluation reveals hidden failures in language-model calculations for environmental science" (under review). Each problem provides only the problem statement (textbook-supplied formulas are stripped); the model writes a Python solve() func...

Agent4cs: A Multi-agent System for Code Summarization in Large Hierarchical Codebases

RADIANT-PET: Reasoning-Augmented PET/CT Lesion Segmentation with Large Language Models and Reinforcement Learning