ENTLORE: A Graph-Grounded Benchmark for Latent Organizational Reasoning in Enterprise Question Answering
ENTLORE tests whether models can recover enterprise relationships that are never directly stated in the document set.
The benchmark builds an audited “truth graph” from routine documents, org tables, and operational records, then releases only the anonymized document corpus. It includes 2,341 documents and 907 questions across explicit lookup, cross-source composition, and latent organizational reasoning. Even with gold documents supplied, models left 30.4% of latent questions unanswered, far worse than explicit or compositional cases. HF Daily Papers' note
The benchmark builds an audited “truth graph” from routine documents, org tables, and operational records, then releases only the anonymized document corpus. It includes 2,341 documents and 907 questions across explicit lookup, cross-source composition, and latent organizational reasoning. Even with gold documents supplied, models left 30.4% of latent questions unanswered, far worse than explicit or compositional cases. HF Daily Papers' note
score 4