MPDocBench-Parse - Truncated Text Merging: leaderboard
Rejoining paragraphs that are split across columns or pages into one paragraph; MPDocBench-Parse (Zhou et al., 2026). Normalized edit distance from 0 to 1 against the merged ground truth; lower is better.
Metric: Normalized edit distance (0-1; paragraphs split across columns or pages, matched to the merged ground truth). Source: arxiv.org. Saturation forecast: Estimated already saturated. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 235B A22B | 0.19 | #477 |
| 2 | Gemini 3.1 Pro (Preview) | 0.23 | #91 |
| 3 | Qwen 3.6 Plus | 0.25 | #217 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mpdocbench-parse-truncated-text-merging · How It Works · Data refreshed daily, snapshot 2026-11-12.