Independent Research Archive
Scientific progress is not document production.
Not the document — the transformation of the mechanism that produced it.
A research cycle can be epistemically successful even when it produces no publishable document — provided it improves the mechanism from which future inquiry emerges.
— from The Reinforcement Learning Contour, 2026
An independent scholarly archive across data, artificial intelligence, clinical systems, and institutional governance. Publication and peer-review status is stated for every record.
Latest Research
The most recent verified works. Each record states its publication and peer-review status.
-
PreprintEmpirical StudyGoverned Local Experiment
Bounded Recursive Self-Improvement Under Human Governance: Evidence from a Seven-Cycle Reinforcement Learning Contour Experiment
A bounded seven-cycle scaffolding-level experiment with a fixed foundation model and separate human promotion authority. Four states were promoted; one failed transition and two NO_CHANGE outcomes were preserved. The public record does not claim AGI, unbounded improvement, autonomous promotion, or production validation.
-
PreprintConceptual Framework
The Reinforcement Learning Contour: From Document Production to Governed Epistemic Transformation in AI-Assisted Scientific Inquiry
Generative artificial intelligence has sharply reduced the technical cost of producing scientific text. Increased production capacity does not, however, entail cumulative knowledge, conceptual novelty, or epistemic progress. This paper introduces the Reinforcement Learning Contour (RLC) as a conceptual and architectural framework for shifting AI-assisted research from repetitive content generation toward governed recursive learning… Version note: current registry metadata is Version 2; the archived PDF and this summary preserve Version 1.
-
PreprintConceptual Framework
When Local Truth Produces Global Falsehood: Moral Grounding, Institutional Volition, and Systemic Error in Artificial Intelligence
Evaluation of artificial intelligence concentrates on local performance — accuracy, calibration, robustness — while the upstream normative and volitional architecture that fixes what a system is for remains largely unexamined by those metrics. Existing scholarship has established that bias, proxy…
-
PreprintEmpirical StudySynthetic Controlled Pilot
Measuring What Survives: Pressure-Filtered Persistence and a Masking Index for LLM-Agent Safety Evaluation
When a safety intervention lowers an agent’s attack-success rate, two different things may have happened: the intervention removed a load-bearing failure route, or the agent learned the measured attacks while the route survived. Frequency metrics are necessary, but they do not by themselves…
-
PreprintEmpirical StudySynthetic Controlled Pilot
A Topological Robustness and Anti-Fragility Framework for LLM Agents
Adversarial evaluation of large-language-model agents is usually expressed as event counts: attack-success rate, policy-violation rate, refusal-failure rate, pass percentage, or robustness score. These measures are necessary, but they do not show whether a safety intervention removed a load-bearing…
-
PreprintConceptual Framework
Human Digital Twins and Self-Knowledge: A Governed Attractor Architecture for Sound Self-Abduction
Large language models embedded in person-modelling systems, and human beings reflecting on themselves, share a structural vulnerability: both can generate fluent, narratively persuasive, yet abductively unsound accounts of a person. Building on a governed abductive architecture developed for…
-
PreprintConceptual Framework
From Plausible Narrative to Sound Abduction: A Governed Abductive Architecture for Medical Digital Twins in Multiple Sclerosis Care
Large language models (LLMs) embedded in medical Patient Digital Twins (PDTs) exhibit a systemic vulnerability: they can generate fluent, narratively persuasive, yet abductively unsound clinical explanations. In this context, abductive soundness means that an explanation preserves mechanistic…
-
PreprintEmpirical Study
TotalSpineSeg: Robust Spine Segmentation with Landmark-Based Labeling in MRI
Accurate segmentation and labeling of spinal structures are essential for diagnosing and managing spinal pathologies. TotalSpineSeg uses cascaded nnU-Net models and landmark-guided iterative labeling to segment vertebrae, intervertebral discs, spinal cord, and spinal canal across MRI contrasts.
Research Areas
- Runtime governance, provenance and execution boundaries for agentic AI.
- Measurement frameworks for adversarial robustness of LLM agents.
- Governed abductive architectures for patient and person modelling.
- Normative and volitional architecture upstream of AI systems.
- Clinical research and imaging/data methods in multiple sclerosis care.
- Lineage, contracts and quality controls that trustworthy AI depends on.