CITATION.cff

cff-version: 1.2.0
message: "If you use this work, please cite it as below."
title: "Proportional Intelligence"
abstract: >-
  A per-instance cost model for difficulty-bound AI. Compute and confidence
  should scale with the difficulty of the single question, not with model size.
  Introduces the per-instance Resource Triedro R_x(B, D, V) with an additive
  conservation law, and the measurable "verification locality" L = d_min(x),
  yielding the falsifiable prediction q* ~ m^(1 - 1/L) for the number of
  spot-checks. Includes a runnable, dependency-free experimental harness
  (synthetic and real-data, HotpotQA-based) and honest scope limits.
authors:
  - family-names: Wavy
    alias: condorstark
type: software
license: MIT
repository-code: "https://github.com/condorstark/proportional-intelligence"
keywords:
  - proportional-intelligence
  - test-time-compute
  - verification-locality
  - hallucination-detection
  - resource-triedro
  - freivalds-algorithm
  - semantic-entropy
year: 2026
date-released: "2026-06-28"
version: "0.1.0"