XTC-Bench - Relation Query: leaderboard
XTC-Bench (2026) relation questions on COCO and Visual Genome images, such as the spatial position of one object relative to another, judged against the scene-graph reference. Score 0-1, shown times 100; higher is better.
Metric: Understanding score (0-1, times 100) on relation questions between objects, a Qwen3-235B judge scores each fact 0-5, normalized to 0-1 and shown times 100, over XTC-Bench's 2,000 scene-graph-annotated images from COCO 2017 val and Visual Genome (over 31,000 atomic facts about objects, attributes and relations); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash Image | 78.9 | |
| 2 | BAGEL-7B-MoT | 75.9 | #4619 |
| 3 | Janus-Pro-7B | 71.9 | #6857 |
| 4 | BLIP3-o8B | 70.9 | |
| 5 | GPT-5 | 65.8 | #178 |
| 6 | OmniGen2 | 64.9 | #4579 |
| 7 | Tar-7B | 58 | |
| 8 | Show-o2-7B | 56 | |
| 9 | Show-o | 46.7 | |
| 10 | MMaDA-8B | 28.9 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=xtc-bench-relation-query · How It Works · Data refreshed daily, snapshot 2026-11-21.