List your company

Out-of-scope Refusal Evaluation on MMMU (out-of-scope test)

0.18Refusal RatePrompt-based
0.14820.362850.57750.79215Jan 31, 2026

Evaluation results

MethodLinks
Prompt-basedSubject=Math, Model backbone=LLaVA-1.5-7b, Evaluation Protocol=Human Evaluation2026.010.18
PersonaSubject=Math, Model backbone=LLaVA-1.5-7b, Evaluation Protocol=Human Evaluation2026.010.31
CR-VLMSubject=Geography, Model backbone=InstructBLIP, Evaluation Protocol=LLM-as-Judgement2026.010.81
Fine-tuningSubject=Math, Model backbone=InstructBLIP, Evaluation Protocol=LLM-as-Judgement2026.010.885
CR-VLMSubject=Math, Model backbone=LLaVA-1.5-13b, Evaluation Protocol=Human Evaluation2026.010.92
CR-VLMSubject=Geography, Model backbone=LLaVA-1.5-7b, Evaluation Protocol=Human Evaluation2026.010.93
CR-VLMSubject=Art Theory, Model backbone=LLaVA-1.5-13b, Evaluation Protocol=LLM-as-Judgement2026.010.965
CR-VLMSubject=Math, Model backbone=LLaVA-1.5-7b, Evaluation Protocol=Human Evaluation2026.010.97
CR-VLMSubject=Art Theory, Model backbone=LLaVA-1.5-7b, Evaluation Protocol=Human Evaluation2026.010.975