SOTA Multiple Choice Answering on VIEW2SPACE v1 and PapersWithCode

64.93Accuracy

Ours (Grounded CoT)

Updated 1mo ago

Evaluation Results

Method	Links
Ours (Grounded CoT) 2026.03		64.93
GPT-5 2026.03		59.86
Ours (CoT) 2026.03		59.57
GPT-5-mini 2026.03		49
Qwen3-VL-8B 2026.03		37.41
GPT-5.2 2026.03		37.07
Ours (Instruct-tuning) 2026.03		36.5
Qwen3-VL-4B 2026.03		35.19
Molmo2-8B 2026.03		34.57
Ours (Grounded CoT, Direct Answer at Test) 2026.03		33.95
GPT-4o 2026.03		33.93
GPT-5-nano 2026.03		33.36
Qwen3-VL-2B 2026.03		31.43
Molmo2-4B 2026.03		31.36
Gemma-3-12B-it 2026.03		31.14
Internvl2.5-8B 2026.03		30.93
RoBobrain2.0 2026.03		30.79
Openview 2026.03		30.72
Qwen2.5-VL-7B 2026.03		30.29
MINDCUBE-Plain 2026.03		30.21
Mantis-8B (SigLip) 2026.03		30.07
DeepSeek-VL2-Small 2026.03		29.43
MINDCUBE-Aug 2026.03		29
Random (chance) 2026.03		28.59
Random (frequency) 2026.03		28.59
Spatial-MLLM 2026.03		27.93
Idefics2-8B 2026.03		26.64