TabVerse - QA (LaTeX Text): leaderboard
TabVerse is a controlled cross-format table-understanding benchmark of 700 balanced question-table pairs drawn from five TableQA sources, with each table aligned across HTML, Markdown and LaTeX as both structured text and rendered images. This board reports Exact-match accuracy (%) answering table questions from LaTeX table text; higher is better.
Metric: Exact-match accuracy (%) answering table questions from LaTeX table text; higher is better. Source: arxiv.org. Saturation forecast: Around June 2027. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 65 | #144 |
| 2 | GPT-5.2 | 57.29 | #176 |
| 3 | Qwen 3 VL 8B Instruct | 52.14 | #981 |
| 4 | Gemma 3 27B (IT) | 51.29 | #1289 |
| 5 | Gemma 3 12B (IT) | 49 | #1837 |
| 6 | Qwen 3 30B A3B Instruct | 48.43 | #1025 |
| 7 | Qwen 3 VL 30B A3B Instruct | 43.57 | #871 |
| 8 | Qwen 2.5 7B Instruct | 42.71 | #2564 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=tabverse-qa-latex-text · How It Works · Data refreshed daily, snapshot 2026-11-13.