Myung Ho Kim

Myung Ho Kim

orcid

Research Focus & Metrics

Representative Publications

View All Publications
4 June 2026 in Preprints.org
Verified
Preprint
Open access
PDF file
MKMyung Ho Kim
Myung Ho Kim
Retrieval-Augmented Generation (RAG) has become the dominant approach for grounding large language model (LLM) responses in external knowledge. Yet despite its widespread adoption, RAG suffers from a structural accountability deficit: it provides no principled mechanism for declaring what information was absent, tracking how individual passages were selected, or preventing the LLM from silently drawing on parametric knowledge when retrieval fails. This paper introduces Pool-Gated Retrieval (PGR), an architecture that addresses these deficits by enforcing three epistemically distinct stages — Horizon, Warrant, and Commitment — each with a formally specified justificatory role. In PGR, the Pool (Horizon stage) collects candidates at high recall; an LLM-driven Warrant gate evaluates each candidate through explicit function calls, applying a Result Relevance Rule and a Knowledge Gap Recognition Rule; and only warrant-approved passages are Committed as the exclusive context for answer generation. Absence is treated as a first-class epistemic fact: a Knowledge Gap is structurally registered and propagated, rather than silently papered over. We present a reference implementation as an open Python package, demonstrate the architecture on a partial-coverage retrieval scenario, and systematically compare PGR against naive RAG across eight identified failure modes. The results show that PGR eliminates absence hallucination by structural design and produces a complete, auditable provenance record for every answer.

22 July 2025 in PsyArXiv
Verified
Preprint
Open access
MKMyung Ho Kim
Myung Ho Kim
We report the discovery of a meaningful structural convergence among four influential theories of mind—Kahneman's dual-system theory, Friston's predictive processing, Minsky's society of mind, and Clark's extended mind—through their unintentional but systematic emergence in a practical AI agent architecture called Agentic Flow. Originally designed to overcome critical limitations of large language models (LLMs), this architecture consists of five interlocking modules—Retrieval, Cognition, Control, Memory, and Action—organized into a repeatable cognitive loop. Although initially inspired only by Minsky and Clark's frameworks, our structural analysis reveals that Agentic Flow partially mirrors core computational patterns described by each of the four theories, suggesting a limited but notable structural convergence. To evaluate this convergence, we conducted controlled experiments comparing structured agents to baseline LLM-based systems across multi-step, conditional reasoning tasks. The structured agent consistently achieved 95.8% task success and demonstrated robust constraint adherence, whereas the baseline system exhibited a 62.3% success rate under the same conditions. While based on a proprietary implementation, evaluation protocols are made available for verification. These experiments were not conducted merely to demonstrate performance advantages, but to qualitatively reveal how theoretical convergence can arise organically through implementation demands rather than deliberate design.We present PEACE as a descriptive meta-architecture that retrospectively draws attention to recurring computational patterns—such as predictive modeling, associative recall, and error-sensitive control—identified in Agentic Flow. Rather than offering a formal theory or aiming to supplant existing frameworks, PEACE serves as a practical abstraction that surfaces design-level regularities seen across otherwise divergent cognitive architectures. Its value lies not in theoretical unification, but in framing a shared vocabulary for analyzing and constructing cognitive systems shaped by real-world implementation demands.Our findings have implications for cognitive science, artificial intelligence, and philosophy of mind. They suggest that intelligent architectures may evolve toward shared structural patterns, shaped not by theory but by the demands of real-world reasoning under uncertainty. We argue that Agentic Flow may offer a partial instantiation—or structural echo—of Newell’s long-sought "unified theory of cognition," discovered not through abstraction, but through necessity. By ‘partial instantiation,’ we do not imply theoretical alignment but structural features independently anticipated by multiple cognitive theories. We emphasize that the observed convergence reflects architectural motifs rather than theoretical completeness.This paper should be read as a position paper—an exploratory reflection on how implementation can surface latent architectural commonalities across divergent cognitive theories, rather than a claim of theoretical unification.

20 November 2025 in PsyArXiv
Verified
Preprint
Open access
MKMyung Ho Kim
Myung Ho Kim
Hallucination in large language models has traditionally been viewed as a correctible error arising from imperfect training. We propose a fundamental reframing: Hallucination may not be a defect but rather a lawful geometric phenomenon emerging from semantic manifold curvature. By integrating information theory, differential geometry, and empirical analysis, we demonstrate that hallucination rate (h) systematically increases with representational curvature (κ), independent of information density (ρ). Using GPT-2 across five reasoning domains (n=1,500), we find a consistent correlation r(κ,h)=0.83 (p≈0.08), providing preliminary evidence that hallucination arises predictably where meaning-space bends most sharply. This transforms hallucination from mystery to measurement—not an error to suppress, but a quantifiable cost of compressing knowledge into finite linguistic form. We introduce the curvature–hallucination relationship h∝f(κ), unifying the manifold hypothesis in representation learning, information-geometry trade-offs, and distortion under abstraction. This framework enables a shift from error correction to geometric control: monitoring curvature to predict and regulate factual instability. What began as a reliability problem may become the first quantitative law linking representation, compression, and truth—the geometry of understanding itself.All Tier 1 experimental code, datasets, and metrics are openly released at the GitHub repository https://github.com/enkiluv/hallucination_curvature_experiment, inviting the research community to replicate, extend, and test whether this relationship generalizes across architectures, languages, and modalities in forthcoming Tier 2 and Tier 3 studies.

9 October 2025 in PsyArXiv
Verified
Preprint
Open access
MKMyung Ho Kim
Myung Ho Kim
Large language models exhibit intelligence without genuine epistemic understanding, revealing a fundamental philosophical gap: the absence of epistemic architecture. This paper introduces the Structured Cognitive Loop (SCL) as an executable epistemological framework for emergent intelligence.Unlike traditional AI research that asks "what is intelligence?" (ontological), SCL asks "under what conditions does cognition emerge?" (epistemological). Situated within contemporary philosophy of mind and cognitive phenomenology, this framework bridges conceptual philosophy and implementable cognition. Drawing on process philosophy, enactive cognition, and extended mind theory, we reconceptualize intelligence not as a possessed property but as a performed process—a continuous loop of judgment, memory, control, action, and regulation.SCL makes three interrelated contributions. First, it operationalizes philosophical insights into computationally interpretable structures, enabling what we term "executable epistemology"—philosophy as structural experiment. Second, it demonstrates that functional separation within cognitive architecture yields more coherent and interpretable behavior than monolithic prompt-based approaches, with empirical support from controlled agent evaluations. Third, it redefines the measure of intelligence: not representational accuracy but the capacity to reconstruct one's own epistemic state through intentional understanding.This framework has implications across philosophy of mind, epistemology, and artificial intelligence. For philosophy of mind, it offers a new mode of engagement where theories of cognition can be enacted and tested. For AI, it grounds behavioral intelligence in epistemic structure rather than statistical regularity. For epistemology, it suggests that knowledge is best understood not as truth-possession but as continuous structural reconstruction within a phenomenologically coherent loop.We situate SCL within debates on cognitive phenomenology, emergence, normativity, and intentionality, arguing that genuine progress requires not larger models but architectures that structurally realize cognitive science principles.

6 September 2025 in PsyArXiv
Verified
Preprint
Open access
MKMyung Ho Kim
Myung Ho Kim
Hallucination in Large Language Models (LLMs) is often treated as a correctable flaw—an artifact to be minimized through dataset curation, scaling, or architectural refinement. This paper advances a different thesis: hallucination is not an incidental failure but an unavoidable structural property of generative intelligence itself. We ground this claim in information theory and cognitive science, arguing that intelligence in LLMs emerges through generalization, a process fundamentally rooted in lossy compression of vast and heterogeneous training data. Lossy compression necessitates reconstructive inference, where missing or ambiguous details are filled in from learned priors. While this mechanism supports creativity, abstraction, and adaptability, it also unavoidably produces outputs that diverge from verifiable facts, labeled as hallucination.To substantiate this argument, we integrate theoretical modeling with a synthesis of recent empirical findings on LLM hallucination rates, scaling laws, and retrieval-augmented architectures. We analyze cases from both open-domain and high-stakes applications, highlighting contexts where hallucination is functionally beneficial and where it presents critical risks. We further examine parallels with human memory errors, particularly reconstructive recall, to situate LLM behavior within broader frameworks of intelligent systems.From this perspective, attempts to eliminate hallucination may dismantle the very mechanisms enabling generalization and intelligence-like behavior. We conclude by outlining implications for AI safety research, policy design for national-scale sovereign AI initiatives, and deployment strategies in domains such as law, medicine, and national security, where balancing creativity with factual reliability is paramount.

28 January 2026 in TechRxiv
Verified
Preprint
Open access
MKMyung Ho Kim
Myung Ho Kim
We introduce LENS (Learning, Evolution, and Natural Language Synthesis), a modular self-improving architecture for autonomous algorithm discovery. Unlike approaches relying on single-model reasoning or shallow combinations of reinforcement learning and language models, LENS decomposes the discovery process into three specialized modules: an RL strategy layer for meta-level decisions, an LLM synthesis layer for creative generation, and an evolution layer for systematic refinement. This structural decomposition enables a self-expanding action space where LLM-generated algorithms become new actions available to the RL policy, creating a persistent knowledge accumulation loop. Early discoveries enable more sophisticated later ones, overcoming the fundamental limitation of static action spaces found in traditional automated machine learning systems. We evaluate LENS on sorting, knapsack optimization, and the traveling salesman problem. The system autonomously evolved from naïve baselines to expert-like algorithms, independently discovering efficient divide-and-conquer sorting structures and demonstrating structural leaps from greedy heuristics to Dynamic Programming and Branch-and-Bound strategies. Experimental traces reveal emergent behaviors including strategic switching, selective refinement, and progressive structural convergence. LENS demonstrates that intelligence can emerge from architectural composition rather than scale alone, enabling systems that invent, test, and improve algorithms without human intervention.