Khondo (PolyInt) - Page Typing: leaderboard
Khondo (2026) page-level domain typing on the PolyInt variant of Bangladeshi government form packets (forms of two or more domains interleaved round-robin page by page): the share of pages given their correct administrative domain. Accuracy in percent, higher is better.
Metric: Page accuracy (%; share of pages assigned their correct administrative domain; forms of two or more domains interleaved round-robin page by page; 210 test packets of 5 to 20 page images of Bangladeshi government forms (Bangla and English, 14 domains); zero-shot order-aware prompt, all page images in one request, JSON output, temperature 1.0, reasoning disabled, invalid outputs retried up to three times). Source: arxiv.org. Saturation forecast: Around February 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.6 Plus (Non-reasoning) | 50.2 | #222 (Qwen 3.6 Plus) |
| 2 | Gemini 3.5 Flash (Minimal) | 46.1 | #101 (Gemini 3.5 Flash) |
| 3 | GPT-5.4 (Non-reasoning) | 41.5 | #123 (GPT-5.4) |
| 4 | Kimi K2.5 (Non-reasoning) | 28.1 | #256 (Kimi K2.5) |
| 5 | GLM-4.6V (Non-reasoning) | 16.4 | #701 (GLM-4.6V) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=khondo-polyint-page-typing · How It Works · Data refreshed daily, snapshot 2026-11-13.