SciDiagramEdit - Semantic Faithfulness: leaderboard
SciDiagramEdit measures instruction-driven editing of scientific figures reconstructed from real paper revision pairs; this board reports blind pairwise semantic-faithfulness win rate for the unassisted model backbones.
Metric: No-skill pairwise win rate (%). Source: arxiv.org. Saturation forecast: Around January 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 74.5 | #31 |
| 2 | GPT-5.4 | 70 | #79 |
| 3 | GPT-5.3 | 65.9 | #103 |
| 4 | GPT-5.1 | 22.4 | #135 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=scidiagramedit-semantic-faithfulness · How It Works · Data refreshed daily, snapshot 2026-10-20.