WildHandBench - Tables: leaderboard
The table part of WildHandBench (2026): recognition of real-world handwritten tables, scored by TEDS, the tree-edit-distance similarity of the recognized table structure and content, in percent; higher is better.
Metric: TEDS (%; tree-edit-distance similarity of the recognized table structure and content over the table documents of 500 real handwritten documents (free text, tables and formulas) in four languages and nine real-world scenarios such as notes, forms, whiteboards and exam sheets). Source: arxiv.org. Saturation forecast: Around February 2028. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 60.44 | #94 |
| 2 | Kimi K3 | 58.99 | #54 |
| 3 | Qwen 3.5 Plus | 57.68 | #232 |
| 4 | Gemini 3.5 Flash | 53.63 | #101 |
| 5 | Claude Opus 4.8 | 52.27 | #56 |
| 6 | Qwen 3 VL 235B A22B (Thinking) | 51.81 | #451 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=wildhandbench-tables · How It Works · Data refreshed daily, snapshot 2026-11-13.