List your company

Volumetric Computed Tomography Multiple Choice Question Answering on CT-RATE MCQ (test)

59.96Existence AccuracyTIF-GRPO w/ Acc Reward (CT-RATE)
46.367249.896153.42556.9539May 19, 2026

Evaluation results

MethodLinks
TIF-GRPO w/ Acc Reward (CT-RATE)#Params=4B2026.0559.9673.8489.3770.2566.65
GRPO w/ Acc Reward (CT-RATE)#Params=4B2026.0555.037388.4167.2963.59
TIF-GRPO w/ Acc Reward (AMOS-MM)#Params=4B2026.0555.0368.3585.9965.5369.67
GRPO w/ Acc Reward (AMOS-MM)#Params=4B2026.0554.1861.1873.9160.4863.91
GPT-5#Params=*2026.0553.5356.9683.5761.2560.54
Qwen3-VL-8B#Params=8B2026.0552.8946.8461.3553.2450.63
Fleming-VL-8B#Params=8B2026.0552.8955.2765.756.4253.92
Hulu-Med#Params=7B2026.0552.1858.7681.0960.4657.03
RadFM#Params=14B2026.0551.8228.6926.0939.9639.53
InternVL3.5-8B#Params=8B2026.0551.1845.1563.7752.4749.18
CT-CHAT#Params=8B2026.0549.2555.2766.6754.7754.22
M3D-LaMed#Params=7B2026.0547.5453.1663.2952.5848.72
Gemini-3-Pro#Params=*2026.0546.8956.5480.1956.9756.7