Skip to content
View sunghunkwag's full-sized avatar

    Block or report sunghunkwag

    Block user

    Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

    You must be logged in to block users.

    Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
    Report abuse

    Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

    Report abuse
    sunghunkwag/README.md

    Sung Hun Kwag

    Independent AI Systems Researcher

    I study where learning systems break down under constrained compute, weak evaluators, distribution shift, and self-modification pressure. I turn those breakdown points into mechanisms for search, validation, memory, and generalization.

    Failure is treated as signal, not noise. Most experiments are designed to be reproducible on CPU.

    Open questions

    RSI stability Can self-improvement loops remain stable under non-leaking episodic memory, rollback constraints, and validation-only gates?

    Related repositories: rsi-metaforge-core, self-improving-research-kernel, DeepNeural-AutoExploration

    Attention-free sequence modeling Can long-range dependencies be handled without attention, softmax, or transformer-style assumptions?

    Related repositories: attention-free-sequence-model, RSI-NAS-Attention-Free, DHC-SSM-Enhanced, SSM-MetaRL-TestCompute

    Parameter-free structure Can representation and memory be built without neural networks, learned weights, or standard gradient-based training?

    Related repositories: field-interference-network, structural-memory-field, OMEGA-THDSE

    Measuring self-improvement Which axes matter for evaluating self-improvement: self-modification depth, operator discovery, meta-adaptation, rollback stability, evaluator robustness, and failure containment?

    Related repository: rsi-bench

    Evaluation discipline

    Experiments use sealed or hidden evaluations where possible, validation-gated synthesis, rollback constraints, evaluator evolution, and failure-to-rule compression.

    Benchmark leakage and evaluator gaming are treated as default threats, not afterthoughts.

    Only changes that pass validation are kept. Failed changes remain as records.

    Contact

    sunghunkwag@gmail.com

    Pinned Loading

    1. ast-grammar-induction-prototypeast-grammar-induction-prototypePublic

      A single-file recursive self-improvement engine that evolves Python programs through AST analysis and statistical grammar learning (EDA).

      Python 1

    2. rsi-metaforge-corersi-metaforge-corePublic

      Experimental Python runtime for validation-gated program synthesis and adaptive search: multi-level meta-learning (meta-meta loops), analogical transfer, grammar-mediated expansion, anti-cheat veri…

      Python 1

    3. MetaRL-Agent-FrameworkMetaRL-Agent-FrameworkPublic

      MetaRL Agent Framework: Modular meta-reinforcement learning system with extensible agent coordination and adaptation.

      Python 1

    4. DeepNeural-AutoExplorationDeepNeural-AutoExplorationPublic

      Recursive Self-Improving Deep Neural Network Autonomous Exploration Algorithm (RSI-DNAX)

      Python