List your company

Spatial relationship understanding on VSR zeroshot

69.25AccuracyGemma3-27B-IT
51.143655.844360.54565.2457Jun 15, 2026

Evaluation results

MethodLinks
Gemma3-27B-ITModel scale=27B2026.0669.2570.7
Thinking with Grounding BoxModel scale=4B, Grounding interface=Box, Grounding reward=w/2026.0668.0886.91
Gemma3-12B-ITModel scale=12B2026.0667.9869.56
Thinking with Grounding BoxModel scale=4B, Grounding interface=Box, Grounding reward=w/o2026.0666.8287.64
Thinking with Grounding PointModel scale=4B, Grounding interface=Point, Grounding reward=w/o2026.0665.3883.88
Thinking with Grounding PointModel scale=4B, Grounding interface=Point, Grounding reward=w/2026.0664.6781.42
Gemma3-4B-ITModel scale=4B2026.0656.6557.94
Non-grounded ThinkingModel scale=4B, Grounding interface=None2026.0651.8479.13