Doc2DB-Bench - Relation Tables (Oracle Entities): leaderboard

Metric: Relation-level cell F1 (%; relationship tables predicted from the document, the schema and the ground-truth entity tables (oracle entity setting); 203 synthesized long-document instances over 42 database schemas from BIRD and Spider in seven domain groups; identical prompts, greedy decoding at temperature 0; cells aligned by global maximum-weight tuple matching, a cell matching on exact numeric equality or at least 90 percent string similarity). Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.

Top models

#ModelScore
1GPT-5.472.33
2Gemini 2.5 Pro65.46
3Claude Opus 4.663.17
4Gemini 2.5 Flash59.5
5Qwen 3 Max55.96
6GPT-4o50.68
7DeepSeek V4 Flash48.93
8Qwen 2.5 72B Instruct31.98
9Qwen 2.5 14B Instruct28.92
10Llama 3.1 70B Instruct16.97

Interactive version: theaggregate.ai/benchmark?slug=doc2db-bench-relation-tables-oracle-entities · How It Works · Data refreshed daily, snapshot 2026-09-29.