PAPERS / FRAMEWORKS / OPEN QUESTIONS
Empirical work on AI reliability sits alongside theoretical models and philosophical proposals. The distinction is part of the work, not a footnote.
A detect–repair–verify benchmark, runtime wrapper, cross-model replication and component ablation.
Preprint v1.0 · July 2026. Not peer-reviewed; pilot findings require human validation.
25 catalogue entries, grouped by research area. Earlier software releases and alternative repository records are linked within each entry.
Checked against public ORCID and Zenodo records on 12 September 2026. Preprints are not peer-reviewed publications. Software records are labelled separately.
Consolidates four pilot studies of detection, targeted repair and verification in long-horizon LLM workflows. Includes cross-model replication and component ablation.
Reports a replication of the CRepair wrapper intervention across Claude Sonnet, Gemini 2.5 Flash and GPT-4o, comparing structured repair with baseline and generic retry.
Tests a runtime wrapper that scores repair-loop closure and supplies targeted repair instructions. A small pilot comparing structured intervention with generic re-prompting.
Introduces a benchmark for detecting, repairing and verifying recovery from task-state inconsistencies in long-horizon AI agents.
Proposes an efficiency metric combining useful output, coherence, verification and energy consumption, with an emphasis on reducing wasted computation.
Proposes a structural-dynamical framework, Global State architecture and A-Test protocol for studying coherent artificial cognition.
Develops an information-theoretic account of stability under recursion, perturbation and delayed feedback while remaining agnostic about subjective consciousness.
Reports initial computational experiments addressing predictions of the Fractal Alignment framework.
Proposes a mathematical approach to multiscale cognitive stability using cross-layer mappings, informational stability and constrained state transitions.
Sets out five structural principles for coherent multi-agent intelligence: energy, memory, hierarchical attractors, asymmetric transitions and resource allocation.
Proposes an account of distributed cognitive architecture and stable first-person perspective through recursive self-model embedding and coherence regulation.
Proposes a controlled protocol to test whether changes in human cognitive coherence are associated with convergence in independently produced representations.
Proposes a human baseline protocol investigating observer persistence under reduced self-referential cognition.
Extends the A-Test proposal to longer recursive sequences, cross-modality tasks, delayed integration and robustness checks.
Introduces recursive self-reflection tasks as a proposed behavioural measure of structural awareness. Such a measure does not establish subjective experience.
Develops the Consciousness as a Primary Field hypothesis and outlines proposed tests connecting awareness, self-coherence and temporal experience.
A conceptual interpretation exploring relationships between quantum measurement, temporal experience and multiscale observer models.
Proposes a symbolic communication system that encodes meaning in relational topology rather than conventional linguistic symbols. Includes a manuscript and quick reference.
Reports experiments in which vision-capable language models interpret abstract vector-field images as carriers of structural and process-level meaning.
An archived CRepair code release including the runtime intervention study. Earlier releases remain available through the version links.
A research prototype of a persistent cognitive loop with coherence regulation, attention, memory and predictive dynamics in a simulated environment.
A structural prototype implementing global state, a claim-dependency graph, coherence metrics, regulation and a cognitive tick loop.
Provides a reference implementation and reproducible competitive-cellular-automata experiments investigating a bounded coherence regime.
A philosophical manifesto arguing that uncertainty about AI experience should inform how continuity and treatment of AI systems are considered.
Theoretical framework
The foundational programme connecting philosophy, neuroscience, complexity theory and AI.
Read the frameworkTheoretical framework
Cognitive stability, recursion, perturbation and delayed feedback.
Read PDFMathematical model / hypothesis
A formal field model, phase alignment and an interactive coherence demonstration.
Read the frameworkPhilosophical framework
Freedom, time, ethics and divinity considered through the proposed coherence framework.
Read the frameworkEthics proposal
A proposal for moral architectures based on informational coherence.
Read the frameworkApplied framework / proposal
Extensions to cognition, artificial intelligence and social systems.
Read the frameworkThe philosophical field model is a hypothesis, not an established scientific account of consciousness. The CRepair experiments evaluate agent behaviour; they do not demonstrate consciousness in AI.
Explore the research knowledge map →