Benchmark RadarAI BENCHMARK PROFILE
CSVFidelity-Bench
CSVFidelity-Bench evaluates knowledge graph construction from statistical tables, focusing on the effect of extraction schema and format coupling on fidelity. It includes 15 datasets, 11 Type-II and 4 Type-III tables, with 1,892 gold standard facts across 6 domains.
- Released
- 2026-05-21
- Readiness
- Inspectable
- Primary field
- General AI
Why it matters
Addresses the gap in evaluating knowledge graph construction fidelity, particularly the overlooked interaction between serialization format and schema constraints, which can cause catastrophic coverage loss.
Motivation
An extraction schema should not reduce knowledge graph fidelity.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.