CSCT: Clock-Selected Compression Theory

Why cognition requires bounds

This site accompanies the CSCT manuscript “Intelligence within Bounds: Why Cognition Requires a Closed Convex Hull”.

CSCT (Clock-Selected Compression Theory) frames cognition as inference constrained by an operational bound: the system can only reliably infer within the region supported by anchored evidence (a closed convex hull of grounded primitives). Outside that bound, outputs are unconstrained and become structurally prone to hallucination.

What you can find here

• Paper source: LaTeX and figure outputs in paper/
• Experiments: high-level summaries and key figures
• How to reproduce: what to run (add your public repo links when ready)

Pages

• Experiments overview
• Metrics & reporting
• Reproduction checklist

Core claim (operational)

• Operational reporting criterion: each experiment defines a success threshold (e.g., similarity ≥ 0.90) for reporting pass/fail.
• Statistical analysis: hypothesis tests are performed on continuous metrics (not the thresholded labels), unless explicitly stated.

Links

• Code: GitHub
• Preprint: 10.5281/zenodo.18408862

Citation

@misc{higuchi2026csct,
  author       = {Higuchi, Naoki},
  title        = {Intelligence within Bounds: Why Cognition Requires a Closed Convex Hull},
  year         = {2026},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.18408862},
  url          = {https://doi.org/10.5281/zenodo.18408862}
}