List your company

Multimodal Reasoning on MathVision (Acc, AUROC, ECE)

28.2AccuracySC
17.69620.42323.1525.877Jul 2, 2026

Evaluation results

MethodLinks
SCBackbone=Qwen2.5VL-7B-Instruct, Averaged Over Seeds=false2026.0728.27019.3
C3RL (Ours)Backbone=Qwen2.5VL-7B-Instruct, Averaged Over Seeds=true2026.0727.962.729
C3RL w/o RefBackbone=Qwen2.5VL-7B-Instruct, Averaged Over Seeds=true2026.0727.360.738.2
RLCRBackbone=Qwen2.5VL-7B-Instruct, Averaged Over Seeds=true2026.0725.462.625.3
RLVRBackbone=Qwen2.5VL-7B-Instruct, Averaged Over Seeds=true2026.0724.651.273.8
BaseBackbone=Qwen2.5VL-7B-Instruct, Averaged Over Seeds=true2026.0724.250.868
SaySelfBackbone=Qwen2.5VL-7B-Instruct, Averaged Over Seeds=true2026.0718.469.413.3
SFT+RefBackbone=Qwen2.5VL-7B-Instruct, Averaged Over Seeds=true2026.0718.161.38.3