thebestai.co encyclopedia

Method

Source Policy & Citation Standards

This document establishes the authoritative sourcing policy, verification rules, and bibliographic citation standards for The AI / SI Encyclopedia (thebestai.co).

#1. The Source-First Mandate

Articles must be constructed from an identifiable body of authoritative evidence. Contributors must never draft narrative assertions from unassisted recall and then attempt to match citations after the fact.

The mandatory workflow is:

  1. Formulate the technical, historical, or empirical research questions.
  2. Retrieve the primary literature and authoritative empirical datasets.
  3. Extract explicit claims, mathematical formulations, and experimental parameters.
  4. Synthesize the findings into the encyclopedia entry.
  5. Verify that each citation directly and explicitly supports the text.

#2. The Four-Tier Source Hierarchy

Tier Category Acceptable Venues & Formats Permissible Uses in Encyclopedia
Tier 1 (Highest) Canonical Academic Literature & Primary Science Peer-reviewed journal publications (Nature, Science, IEEE TPAMI, JMLR); top-tier peer-reviewed proceedings (NeurIPS, ICML, ICLR, ACL, CVPR); seminal textbooks (Russell & Norvig, Bishop, Goodfellow et al., Sutton & Barto). Absolute authority for mathematical proofs, core algorithmic definitions, foundational history, and verified experimental findings.
Tier 2 Frontier Lab Technical Reports & Official Filings Official research technical reports from frontier labs (Google DeepMind, OpenAI, Anthropic, Meta AI, FAIR) with transparent methodologies; government statutes (EU AI Act, US Executive Orders); official standards (ISO/IEC, NIST). Authoritative for model architecture disclosures, specific benchmark results reported by developers, and statutory legal definitions.
Tier 3 Non-Peer-Reviewed Preprints Preprints deposited on arXiv or open research archives from recognized research groups. Permissible for frontier results under active study. Must be explicitly cited as preprints and classified as Preliminary evidence until independently replicated.
Tier 4 Investigative Journalism & Industry Documentation Major archival records, reputable technology publications (MIT Technology Review, IEEE Spectrum, Reuters), documented corporate filings (10-K, SEC). Permissible for historical corporate milestones, funding rounds, antitrust cases, and facility investments.
Unacceptable Unverified / Ephemeral Content Promotional blog posts lacking methodology, sponsored marketing articles, social media rumors, forum commentary, AI chatbot responses. Strictly prohibited. Never cite promotional press releases as proof of scientific or technical capability.

#3. Citation Formatting Requirements

Every cited source in Section 11 of an article must provide:

  • Author(s): Full surname and initials for all primary contributors (or lab name for collective institutional reports).
  • Year: Year of original publication or official public archive deposit.
  • Title: Exact title of the paper, book, or technical report.
  • Venue: Journal, conference proceedings, or publishing house.
  • Identifier: Persistent URL, DOI, or arXiv ID (https://doi.org/... or arXiv:XXXX.XXXXX).
  • Annotation: A concise 1-sentence note stating exactly what mathematical formulation, historical event, or empirical result this citation substantiates.

#Example Formats:

  • Seminal Conference Paper:

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention Is All You Need. Advances in Neural Information Processing Systems (NeurIPS 2017), 30, 5998–6008. https://arxiv.org/abs/1706.03762. [Seminal publication introducing the Transformer architecture, multi-head self-attention, and sinusoidal positional encodings].

  • Foundational Textbook:

    Russell, S., & Norvig, P. (2020). Artificial Intelligence: A Modern Approach (4th ed.). Pearson. ISBN: 978-0134610993. [Canonical reference for definitions of rational agents, search algorithms, and knowledge representation].

  • Frontier Technical Report:

    DeepMind. (2020). AlphaFold: A Solution to a 50-Year-Old Grand Challenge in Biology. DeepMind Research Blog & CASP14 Proceedings. https://doi.org/10.1038/s41586-021-03819-2. [Documented breakthrough in computational structural biology and protein structure prediction].

#4. Zero Tolerance for Citation Hallucination

Gemini CLI and all contributing editors operate under a strict zero-tolerance policy regarding citation hallucination:

  1. No Invented DOIs: Every DOI must resolve to the real publication.
  2. No Imagined Collaborators: Do not guess co-authors. If uncertain, verify against the published paper header.
  3. No Phantom Benchmarks: Never cite a synthetic benchmark score that cannot be verified in the primary report.
  4. Unresolved Citation Marking: If a primary paper cannot be verified during an editing session, mark the claim with [VERIFICATION NEEDED: Source unverified] rather than fabricating a citation.