Benchmarking Complex Multimodal Document Processing Pipelines: A Unified Evaluation Framework for Enterprise AI

TL;DR

This paper introduces EnterpriseDocBench, a unified evaluation framework for enterprise multimodal document processing pipelines, assessing parsing, retrieval, and generation quality on a common corpus.

Contribution

It presents a comprehensive benchmarking approach, including a new corpus and evaluation metrics, for assessing entire enterprise document AI pipelines holistically.

Findings

Hybrid retrieval slightly outperforms BM25 in relevance.

Hallucination rates vary with document length, peaking at short and very long documents.

Weak correlations between pipeline stages suggest limited cascading effects.

Abstract

Most enterprise document AI today is a pipeline. Parse, index, retrieve, generate. Each of those stages has been studied to death on its own -- what's still hard is evaluating the system as a whole. We built EnterpriseDocBench to take a swing at it: parsing fidelity, indexing efficiency, retrieval relevance, and generation groundedness, all on the same corpus. The corpus is built from public, permissively licensed documents across six enterprise domains (five represented in the current pilot). We ran three pipelines through it -- BM25, dense embedding, and a hybrid -- all with the same GPT-5 generator. The headline numbers: hybrid retrieval narrowly beats BM25 (nDCG@5 of 0.92 vs. 0.91), and both beat dense embedding (0.83). Hallucination doesn't grow monotonically with document length -- short documents and very long ones both hallucinate more than medium ones (28.1% and 23.8% vs. 9.2%). Cross-stage correlations are very weak: parsing->retrieval r=0.14, parsing->generation r=0.17, retrieval->generation 0.02. If quality were cascading the way most of us assume, those numbers would be much higher; they aren't. Design caveats are real (parsing fixed, generator shared, automated proxy metrics) and we don't oversell the result. One result that genuinely surprised us: factual accuracy on stated claims is 85.5%, but answer completeness averages 0.40. The system is right when it answers -- it just leaves things out. That gap matters more for real deployments than the headline accuracy number does. We also describe three reference architectures (ColPali, ColQwen2, agentic complexity-based routing) which are not yet integrated end-to-end. Framework, metrics, baselines, and collection scripts will be released open-source on acceptance.