⚙ How is this computed? Methods: Knowledge & RAG
Knowledge Graph · Deterministic evidence search

Search the LACA Evidence Index

Search and rank curated records from GenePedia, LiteraturePedia, ScholarPedia, and DatabasePedia. Inspect the retrieval score, matched terms, evidence excerpt, and source link for every result.

Bilingual search Field-weighted ranking Matched terms + scores No LLM · Deterministic
Search evidence → How it works →
222 Curated records indexed
15 Longevity genes
131 PubMed-linked papers
76 Scholars and resources

Evidence search workbench

Search in English or Chinese. Exact identifiers and names receive the highest weights; records are ordered by a deterministic field-weighted score.

NO LLM · DETERMINISTIC Checking index…
Search in

Search the curated evidence index

Results show rank, retrieval score, match reason, matched terms, evidence excerpt, and source link. No language model generates or reranks these records.

Ranking matching records…

Ranked evidence records

0 ranked records
NO LLM · DETERMINISTIC

Deterministic evidence digest

Extracted from the ranked records
    Indexed evidence substrate

    Coverage of the first four Knowledge Graph hubs

    The current snapshot contains structured records already published in LACA.

    GenePedia · 15 Symbols, names, pathways, PMID links, journals, years, and page evidence scores. LiteraturePedia · 131 PMID/DOI records, titles, authors, journals, topics, keywords, and affiliations. ScholarPedia · 28 Researchers, institutions, fields, topics, and clearly marked metric snapshots. DatabasePedia · 48 Resource descriptions, types, categories, organisations, features, status, and URLs.
    Retrieval contract

    How Evidence Search works

    This tool prioritizes auditability: it uses field-weighted lexical retrieval and deterministic extraction. It does not call a language model or generate an AI answer.

    Normalize

    Resolve Chinese/English aging terms, gene symbols, PMIDs, resource names, and common aliases.

    Retrieve

    Rank exact identifiers, titles, keywords, metadata, and descriptions across the selected hubs.

    Digest

    Extract a bounded digest directly from retrieved fields; no language model or model memory is used.

    Cite or abstain

    Attach source IDs and inspectable links to every result, or state that evidence was not found.