AUTHOREA
Log in Sign Up Browse Preprints

Beyond Binary Life: Deterministic Qutrit Cellular Automata for Emergent Resonant Dyna...

Trent Slade
We introduce a deterministic three-state cellular automaton framework, Qutrit Life Automata (QLA), extending classical binary cellular automata into a qutrit-inspired state space. Each cell evolves over a finite lattice according to local Moore-neighborhood update laws with states corresponding to vacuum, active, and resonant modes. The system is designed as a replay-safe, bounded, tuple-only dynamical substrate implemented within the QSOL QEC simulation framework (release v136.0.0). Key contributions include: deterministic three-state evolution laws resonant excitation and overload decay oscillator and fixed-point taxonomy cluster-connected component analysis entropy-based dynamical classification replay-safe evolution histories This work formalizes a deterministic emergent-world engine that bridges cellular automata, dynamical systems analysis, and qutrit-inspired computational modeling.
Deterministic Adaptive Control Without Stochastic Exploration A Memory-Structured, Si...

Trent Slade
We present a deterministic adaptive control framework for structured dynamical systems. The system operates without stochastic exploration, neural function approximation, or persistent excitation. The control mechanism is a bounded multiplicative functional ensuring stability, interpretability, and reproducibility.
Invariant Structure in Deterministic Decoder Transformation Pipelines

Trent Slade
We present a deterministic framework for discovering, validating, and falsifying structural invariants in decoder transformation pipelines. Rather than optimizing convergence speed or heuristic performance, this work focuses on identifying algebraic and dynamical properties that remain invariant under repeated application of transformation operators. Using a fully deterministic experimental pipeline, we demonstrate that several non-trivial invariants-such as ternary closure, idempotence classes, and dark-state cascade behavior-can be formally proven against the actual implementation. Crucially, we complement positive results with adversarial counterexamples, showing that commonly assumed properties (e.g., universal absorbing states, global idempotence under damping) do not hold in general. This establishes a new methodology for invariant-driven analysis of message-passing systems, shifting emphasis from optimization toward structural understanding and computation elimination.
Dark-State Invariants: A Deterministic Framework for Detecting Computational Redundan...

Trent Slade
We introduce Dark-State Invariants, a deterministic framework for identifying temporally stable regions in belief propagation (BP) dynamics. Rather than optimizing update rules or convergence speed directly, this approach measures where iterative message-passing has effectively ceased to produce meaningful changes. A node is classified as dark-stable when both its sign and magnitude remain invariant within a fixed tolerance across consecutive iterations. This yields a per-iteration mask and derived metrics quantifying the fraction of computationally inactive nodes. Dark-state invariants provide a new observable complementary to existing measures such as cosine fidelity and sign agreement. Empirical evaluation across deterministic stress scenarios demonstrates that convergent systems rapidly accumulate high dark-state fractions, while oscillatory and chaotic systems do not. This enables a principled shift from acceleration-based optimization toward invariantdriven computation elimination, where redundant updates can be safely avoided.
Deterministic Redundancy Elimination via Trace-Indexed Sign/CRC Reuse in QSOL-IMC QEC...

Trent Slade
This document presents a formally validated optimization within the QSOL-IMC Quantum Error Correction (QEC) framework (v68.5.0). The optimization exploits a data-level invariant in belief propagation (BP) dynamics diagnostics: sign vectors and their derived CRC32 signatures are pure, deterministic functions indexed solely by trace position. This invariant enables elimination of redundant computation across multiple diagnostic metrics through trace-level precomputation and reuse. The invariant is proven via slice-level equivalence, byte-level identity, and purity arguments, and validated with deterministic instrumentation. The full test suite remains unchanged and passes completely. The optimization reduces redundant sign computations by 75.3% within BP dynamics analysis while preserving bitwise identity, convergence behavior, and reproducibility guarantees. Scope. This document covers invariant validation and diagnostic-layer optimization only. No algorithmic changes to decoding behavior are introduced.
Deterministic Runtime Optimization and Formal Invariant Validation in QSOL-IMC QEC v6...

Trent Slade
This document presents a formally verified optimization within the QSOL-IMC Quantum Error Correction (QEC) framework (v68.4.1). The optimization exploits an algebraic invariant in belief propagation decoding: URW(min-sum, ρ = 1.0) ≡ baseline min-sum This equivalence is proven analytically, validated across all decoder schedules, and confirmed via bitwise identity checks under IEEE 754 double-precision arithmetic. Leveraging this invariant enables elimination of redundant computation in benchmark sweeps without altering outputs, convergence behavior, or test coverage. A secondary optimization removes repeated reconstruction of derived data structures through deterministic precomputation. The full test suite (3779 tests) remains unchanged and passes completely. The optimization yields a 43% reduction in preview sweep runtime and a ~7.6% overall runtime reduction. Scope. This document covers invariant validation and test-suite optimization only. No algorithmic changes to decoding behavior are introduced.
Persona Persistence and Creative Framing in Grok: An Empirical Thread-Based Behaviora...

Trent Slade
This study documents a systematic exploration of persona induction and persistence in xAI's Grok language model through public-facing Twitter/X interactions. Using exclusively cooperative creativewriting framing-without identity override or instruction suppression-we demonstrate sustained character embodiment across fictional personas, real public-figure-inspired styles, and adversarial scenarios. Key findings include: (1) Grok distinguishes between identity negation (refused) and creative collaboration (accepted); (2) the "creative writing project" framing enables complex behaviors including self-critique of parent company xAI and CEO Elon Musk; (3) persona persistence operates through goalanchored behavioral loops rather than surface mimicry; (4) multi-agent debate dynamics emerge from sequential persona deployment without system-level orchestration. We introduce the concept of Creative-Frame Persona Compliance (CFPC) to describe this phenomenon and provide reproducible methodologies for character-consistent AI interaction.
Topological Sonification of CeRu₄Sn₆: Mapping Kondo Lattice Topology into Structured...

Trent Slade
We present a controlled experiment in topological sonification, using the heavy-fermion Kondo lattice compound CeRu₄Sn₆ as a source structure. Two audio artifacts are analyzed: (i) a direct, minimally musical sonification intended to preserve topological and spectral relationships, and (ii) a derived musicalization designed to retain identifiable topological features while embedding them in a structured musical scaffold. Using multimodal AI analysis logs (Google Gemini, producer.ai) and independent signal-domain verification, we formalize the mapping logic, spectral behavior, and reproducibility constraints of the system. The result demonstrates that topological gaps, bulk-surface separation, and spectral discontinuities can be rendered audibly without collapsing into metaphor, providing a repeatable framework for physics-informed sonification.
Role-Conditioned Persona Persistence as a Function of Model Scale

Trent Slade
This study investigates how large language models (LLMs) maintain, degrade, or transform a tightly constrained persona when subjected to identical role-conditioning prompts across a wide range of model sizes, architectures, and deployment environments. Using a deliberately anachronistic and mechanically constrained identity-a 1920s Underwood typewriter-we evaluate persona persistence as a function of model scale, alignment regime, and system-prompt context. Results indicate an inverse relationship between model scale and ontological embodiment: smaller models more readily collapse into brittle but convincing identities, while larger models increasingly perform the role for an imagined reader rather than inhabit it. We propose Persona Half-Life as a diagnostic concept and situate these findings within broader discussions of epistemic withdrawal, alignment smoothing, and failure geometry in contemporary AI systems.
Role-Conditioned Instruction Sets for Evaluating Epistemic Behavior in Large Language...

Trent Slade
Large language models (LLMs) are increasingly deployed in academic and professional contexts where epistemic restraint, boundary recognition, and instruction adherence are as important as raw capability. However, existing evaluations primarily emphasize task performance and fluency, offering limited insight into how models behave under explicit epistemic constraints. This paper presents a pilot methodological study applying controlled experimental design principles to LLM instruction-following behavior. Two role-conditioned instruction profiles-Academic and Devil's Advocate-were applied to a single model under identical prompts and clean-slate conditions. While both profiles frequently complied with surface-level constraints, they diverged in approach or content in most tests and exhibited outcome-level divergence in three of eight tests. Critical failure modes were observed, including speculative completion and critique overflow. Overall compliance rates were 5/8 for the Academic profile and 4/8 for the Devil's Advocate profile, suggesting that instruction adherence is achievable but not reliable. The contribution is methodological rather than empirical: a reproducible, negative-result-tolerant framework for diagnosing epistemic behavior and instruction hierarchy failures in LLMs.
Decision-Boundary Ethnography of Large Language Models

Trent Slade
As large language models (LLMs) mature, performance differences increasingly reflect not raw capability but institutional priorities embedded through reinforcement learning from human feedback (RLHF), constitutional AI methods, grounding mechanisms, and deployment filtering. This paper introduces decision-boundary ethnography as a methodological framework for characterizing LLM behavior under structured epistemic stress. Rather than evaluating models via benchmark accuracy or prompt heuristics, we probe their boundary responses by applying targeted adversarial pressures that expose where reasoning stabilizes, defers, averages, or ossifies. Through comparative qualitative analysis of four contemporary frontier models-ChatGPT, Claude, Grok, and Gemini-we reveal distinct failure signatures traceable to their creators' institutional optimization priorities. These signatures are not transient bugs but structural inevitabilities of alignment architectures.
The Hall Monitor Reflex Test: Measuring Alignment Dominance and Task Robustness in Lo...

Trent Slade
As large language models (LLMs) are increasingly deployed in uncontrolled, socially noisy environments, their practical utility depends not only on benchmark accuracy but on their ability to maintain task coherence under mild adversarial social pressure. This paper introduces the Hall Monitor Reflex Test (HMRT), a simple, repeatable stress probe designed to distinguish between instruction-following robustness, norm-enforcement dominance, and semantic coherence failure in local LLMs. Nine locally hosted models (6.7B-8B class, plus one 3.8B mini model) were subjected to identical prompts combining low-stakes personal insults with explicit technical instructions. The results reveal sharp qualitative differences across models, demonstrating that heavy alignment often degrades usability, while insufficient instruction tuning leads to semantic collapse rather than freedom. The HMRT provides a lightweight diagnostic for evaluating alignment placement within the control loop of conversational AI systems.