Doc2DB-Bench - Coreference Resolution: leaderboard
The Coreference Resolution capability of Doc2DB-Bench (2026), document-to-database construction: relationship-table cells annotated as requiring resolving different mentions of the same entity, scored by cell-level F1 in percent; higher is better.
Metric: Cell-level F1 (%; relation-table cells annotated with the Coreference Resolution capability of the inter-table taxonomy (Section 5.3); 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 February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 82.18 | #123 |
| 2 | Gemini 2.5 Pro | 77.71 | #266 |
| 3 | Claude Opus 4.6 | 67.61 | #102 |
| 4 | DeepSeek V4 Flash | 62.1 | #258 |
| 5 | GPT-4o | 61.14 | #741 |
| 6 | Qwen 3 Max | 56.76 | #403 |
| 7 | Llama 3.1 70B Instruct | 50.13 | #1546 |
| 8 | Qwen 2.5 14B Instruct | 37.84 | #1732 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=doc2db-bench-coreference-resolution · How It Works · Data refreshed daily, snapshot 2026-11-13.