OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Evaluation & Observability

Enterpriserag Bench

Enterpriserag Bench is a dataset, benchmark and dataset-generation codebase for evaluating retrieval-augmented generation systems on simulated company-internal documents. Its core release contains slightly more than 500,000 documents and 500 questions.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Enterpriserag Bench — OpenViST project profile in AI Evaluation & Observability.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when testing how a RAG system handles varied internal-style business records rather than a single public-document collection, or when you need a shared question set for comparison.

Who it is for

  • Teams benchmarking retrieval and answer quality in RAG systems.
  • Researchers and developers evaluating RAG agents on enterprise-style information.
  • Teams creating comparable synthetic document datasets for different industries or company scales.

Documented uses

  • Compare RAG systems against a shared enterprise-document question set.
  • Evaluate performance on semantic, cross-document, conflicting-information and answer-not-found questions.
  • Generate similar synthetic enterprise-document datasets for other industries or company scales.
  • Consult leaderboard rankings for the benchmark.
Before you choose

Requirements & limits

The public GitHub repository specifies an MIT license for its code. The dataset is linked from Hugging Face and repository releases.

  • The simulated corpus represents a company providing AI model inference as a service, with records across nine source types.
  • The 500 core questions span ten categories, from basic lookup to conflicts, completeness, high-level synthesis and information-not-found cases.
  • A separate set of 100 metadata-dependent questions is available but excluded from leaderboard evaluation.
  • The repository provides dataset-generation code for adapting industry and company-scale characteristics.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Download the EnterpriseRAG-Bench dataset and question set

Obtain the benchmark documents and questions for use in a RAG evaluation workflow.

Before you start

  • Access to the project’s GitHub README or its linked Hugging Face dataset.

Steps

  1. Download the dataset from the latest GitHub release or the linked Hugging Face dataset.

  2. Choose either all_documents.zip for the full corpus or source-type slice archives, each containing up to 5,000 documents.

  3. Get the question set from questions.jsonl in the repository or from the releases.

  4. For instructions on using the data or running answer evaluation, consult the project’s Quickstart Guide or answer_evaluation/README.md, respectively.

Expected outcome

The benchmark document archive and question set are available to take into a RAG evaluation workflow.

Guide limitations

  • The README points to separate documentation for running answer evaluation; it does not provide those execution steps here.
  • The core question set has 500 questions; an additional 100 metadata-dependent questions are in extra_questions.jsonl and are excluded from the leaderboard.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
9 October 2026
Original website title
onyx-dot-app/EnterpriseRAG-Bench · Official homepage title · retained in its source language
How we review sourcesSuggest a correction