List your company

Visual Reasoning and Calibration on Aggregate of 12 evaluation benchmarks Average

80.3AccuracyVL-Calibration
64.59668.67372.7576.827Apr 10, 2026

Evaluation results

MethodLinks
VL-CalibrationModel=Qwen3-VL-30B2026.0480.376.78.2
VL-CalibrationModel=InternVL3.5-4B2026.0468.970.110.3
RLCRModel=InternVL3.5-4B2026.0465.664.920.9
P(True)Model=Qwen3-VL-30B2026.0465.256.938.8