cff-version: 1.2.0
message: "If you use this dataset or worksheet, please cite it using the metadata below."
title: "AI Research Assistant Repeatability Benchmark 2026"
type: dataset
version: 1.0.0
date-released: 2026-08-01
authors:
  - name: "AI News & Updates Editorial Team"
abstract: >-
  Fifteen run-level observations from three matched attempts with
  ResearchRabbit, Elicit, Scite, Consensus and Undermind, recording
  repeatability, traceability, access barriers and preserved failures.
keywords:
  - artificial intelligence
  - information retrieval
  - research assistants
  - reproducibility
  - repeatability
repository-code: "https://github.com/dannyrasul/ai-research-assistant-repeatability-benchmark-2026"
url: "https://ainewsandupdates.com/how-to-test-ai-research-tools/"
contact:
  - name: "AI News & Updates Editorial Team"
    email: "hello@ainewsandupdates.com"