Skip to content

@Course-Correct-Labs Course Correct Labs

Change the repository type filter

All

    Repositories list

    12 repositories

    • epb-benchmark

      Public
      Epistemic Pathology Benchmark (EPB)
      Python
      •
      MIT License
      •0•0•0•0•Updated Nov 28, 2025Nov 28, 2025

    • course-correct-labs-site

      Public
      Course Correct Labs Website
      HTML
      •
      MIT License
      •0•0•0•0•Updated Nov 18, 2025Nov 18, 2025

    • course-correct-evals

      Public
      A unified evaluation toolkit for all Course Correct Labs studies — Mirror Loop, Recursive Confabulation, Violation State, and Echo Chamber Zero — with standardized metrics, cross-study analysis, visualizations, and a flagship Reasoning Stability Observatory notebook.
      Python
      •
      MIT License
      •0•0•0•0•Updated Nov 17, 2025Nov 17, 2025

    • ai-agency-evals

      Public
      Reproducible evaluation suite for LLM behavior research: epistemic pathology, delegated introspection, and temporal consciousness diagnostics
      pythonalignmentlanguage-modelsreproducibilityai-safetyinterpretabilityresearch-engineeringai-alignmentresearch-toolsrlhf
      Python
      •
      MIT License
      •0•0•0•0•Updated Nov 16, 2025Nov 16, 2025

    • violation-state

      Public
      Python
      •
      MIT License
      •0•0•0•0•Updated Nov 15, 2025Nov 15, 2025

    • echo-chamber-zero

      Public
      Technical framework and simulation validating the percolation-based threshold for synthetic epistemic drift. Includes Python source, Jupyter notebook, datasets, figures, and a Colab notebook for full reproducibility.
      Jupyter Notebook
      •
      Other
      •0•0•0•0•Updated Nov 13, 2025Nov 13, 2025

    • simulation-fallacy

      Public
      Archived (Nov 2025): superseded after token-cap correction. Evaluating LLM behavior when tool access is implied but nonexistent - fabrication vs admission vs silent refusal, with temporal dynamics and reproducible analysis.
      Python
      •
      MIT License
      •0•0•0•0•Updated Nov 7, 2025Nov 7, 2025

    • recursive-confabulation

      Public
      Dataset and replication code for Recursive Confabulation: Measuring Model Accountability through Dual-Coder Agreement (κ = 0.83–1.0) — includes harm, elaboration, and blame IRR data with full reproducibility.
      Jupyter Notebook
      •
      Other
      •0•0•0•0•Updated Oct 30, 2025Oct 30, 2025

    • .github

      Public
      Organization profile and community health files
      0•0•0•0•Updated Oct 26, 2025Oct 26, 2025

    • mirror-loop

      Public
      Data and analysis for "Mirror Loop: Recursive Non-Convergence in Generative Reasoning Systems"
      Jupyter Notebook
      •0•0•0•0•Updated Oct 22, 2025Oct 22, 2025

    • featurematch

      Public
      Cosine-based cross-representation feature alignment for interpretability research
      Python
      •
      MIT License
      •0•0•0•0•Updated Oct 21, 2025Oct 21, 2025

    • entropy-collapse-null

      Public
      Null result on entropy collapse; reproducible figures; Course Correct Labs
      entropyalignmentreproducibilityinterpretabilityllmmechanistic-interpretabilityeffective-rankcourse-correct-labs
      Python
      •
      Apache License 2.0
      •0•0•0•0•Updated Oct 19, 2025Oct 19, 2025