EL
Eriva Labs
Go back Home

Eriva Lens Indexing Policy

Last Updated: 7 July 2026

Eriva Labs currently focuses exclusively on the Artificial Intelligence and Machine Learning domains. To maintain high standards of relevance, reproducibility, and deep-dive value, we selectively curate and index public academic papers from arXiv.org that fall within these disciplines.

These specific filtering rules are designed to select publications that are highly fit for AI-assisted research and analysis (ensuring there is sufficient text context, technical detail, and reproducible structure for our Large Language Models to generate accurate, high-quality summaries).

This page outlines the exact, objective filtering rules used by our automated ingestion pipeline to index papers in our vector database (LanceDB).

1. Category Filter (Must-Pass Gate)

A paper must belong to at least one of the following primary research categories to be considered for ingestion:

  • cs.AI — Artificial Intelligence
  • cs.LG — Machine Learning
  • cs.CV — Computer Vision and Pattern Recognition
  • cs.CL — Computation and Language (Natural Language Processing)
  • cs.NE — Neural and Evolutionary Computing
  • stat.ML — Machine Learning (Statistics)

Papers outside these core domains (e.g., general physics, biology, pure mathematics) are not pre-indexed.

2. Quality & Curation Triggers

If a paper passes the Category Gate, it must satisfy at least one of the following criteria to be added to the search index:

A. Associated Open-Source Code

The paper's metadata, abstract, or comments must reference a public repository to encourage reproducibility:

  • Includes links to GitHub (github.com) or Hugging Face (huggingface.co).

B. Peer-Reviewed Venues

The paper has been peer-reviewed and accepted by one of the top-tier international AI conferences:

  • NeurIPS, ICLR, ICML, CVPR, ICCV, ECCV, ACL, EMNLP, or AAAI.

C. Registered DOI

The paper has been assigned a Digital Object Identifier (DOI) from an external, legitimate academic publisher or journal, confirming formal publication outside of pre-print servers.

D. Leading Research Institutions

The research is authored by researchers affiliated with top-tier AI labs or academic institutions, including:

  • Industry Labs: DeepMind, OpenAI, Meta AI (FAIR), Google Research, Microsoft Research, Anthropic, NVIDIA, Cohere, Mistral, Stability AI, xAI, etc.
  • Academic & Institutes: Stanford, MIT, UC Berkeley, Carnegie Mellon, Oxford, Cambridge, MILA, Vector Institute, Max Planck, INRIA, CNRS, etc.

3. Format & Page Count Validation

To ensure Eriva Lens only indexes thorough papers (rather than short summaries or slide decks), our ingestion pipeline downloads the PDF file and validates the document formatting:

  • Minimum Page Length: The paper must be at least 5 pages long.
  • Excluded Formats: This rule automatically filters out short extended abstracts, workshop posters, lecture slides, and single-page announcements.

4. Querying Unindexed Papers

If a specific research paper you need is not pre-indexed in our database, it does not mean you cannot analyze it. Eriva Lens provides two ways to run instant AI summaries on any paper:

  1. Custom PDF Upload: Drag and drop any local PDF file directly into Eriva Lens. The paper is parsed private to your session.
  2. Direct arXiv Ingestion: Paste the full arXiv link (e.g., https://arxiv.org/abs/...) to ingest the paper on-demand.