List your company

Vision-Language Question Answering on Pooled MLLM Suite (GQA, LLaVA-Wild, MMMU Pro, MME-Finance) (test)

5.4Expected Calibration Error (ECE)InternalInspector
3.8614.25524.6535.045May 11, 2026

Evaluation results

MethodLinks
InternalInspectorLVLM=Qwen/Qwen3-VL-8B-Instruct, n=30,2412026.055.41773.382.689.779.6
BICRLVLM=llava-hf/llava-v1.6-vicuna-13b-hf, n=30,5012026.055.618.271.478.187.778.9
PIKLVLM=Qwen/Qwen3-VL-8B-Instruct, n=30,2412026.057.51872.281.588.477.2
BICRLVLM=Qwen/Qwen3-VL-8B-Instruct, n=30,2412026.058.917.472.978.990.380.1
SAPLMALVLM=Qwen/Qwen3-VL-8B-Instruct, n=30,2412026.0510.619.472.882.886.474.2
PIKLVLM=llava-hf/llava-v1.6-vicuna-13b-hf, n=30,5012026.0510.819.968.978.287.177.3
PIKLVLM=OpenGVLab/InternVL3_5-14B-HF, n=30,5002026.0510.820.270.580.386.473.8
PELVLM=llava-hf/llava-v1.6-vicuna-13b-hf, n=30,5012026.0512.823.56377.377.368.5
InternalInspectorLVLM=llava-hf/llava-v1.6-vicuna-13b-hf, n=30,5012026.0513.822.365.477.881.471.9
CCPSLVLM=OpenGVLab/InternVL3_5-14B-HF, n=30,5002026.0514.723.867.88071.658.2
CCPSLVLM=llava-hf/llava-v1.6-vicuna-13b-hf, n=30,5012026.0516.32267.478.87672.9
SAPLMALVLM=llava-hf/llava-v1.6-vicuna-13b-hf, n=30,5012026.0516.52367.378.88272.5
SAPLMALVLM=OpenGVLab/InternVL3_5-14B-HF, n=30,5002026.0516.623.369.680.876.265.4
PELVLM=OpenGVLab/InternVL3_5-14B-HF, n=30,5002026.0516.725.966.68060.343.4
PELVLM=Qwen/Qwen3-VL-8B-Instruct, n=30,2412026.0519.82667.380.466.252.9
Self-ProbingLVLM=OpenGVLab/InternVL3_5-14B-HF, n=30,5002026.0521.524.768.480.876.870.8
Self-ProbingLVLM=Qwen/Qwen3-VL-8B-Instruct, n=30,2412026.0524.427.569.781.27759.5
P(True)LVLM=llava-hf/llava-v1.6-vicuna-13b-hf, n=30,5012026.0526.230.945.734.867.954.7
CCPSLVLM=Qwen/Qwen3-VL-8B-Instruct, n=30,2412026.0528.745.653.954.566.244.9
Self-ProbingLVLM=llava-hf/llava-v1.6-vicuna-13b-hf, n=30,5012026.0528.93065.278.181.967.3
P(True)LVLM=OpenGVLab/InternVL3_5-14B-HF, n=30,5002026.0541.241.557.2707859.5
P(True)LVLM=Qwen/Qwen3-VL-8B-Instruct, n=30,2412026.0543.944.254.66776.254.6