List your company

Multimodal Reasoning on Geo3K (accuracy)

45.59AccuracySFT-Oracle
25.341230.598135.85541.1119May 12, 2026May 18, 2026May 25, 2026Jun 1, 2026Jun 8, 2026Jun 15, 2026Jun 22, 2026

Evaluation results

MethodLinks
SFT-OracleBackbone=InternVL3-8B2026.0545.59
VISTA-SFTBackbone=InternVL3-8B2026.0545.59
R3VBackbone=InternVL3-8B2026.0543.76
STaR+Backbone=InternVL3-8B2026.0543.43
CFPOGBackbone=Qwen2.5-VL-7B, Hyperparameters=gamma=0.02 + No Ent2026.0643.18
STaRBackbone=InternVL3-8B2026.0541.93
SFT-SeedBackbone=InternVL3-8B2026.0541.6
GRPOBackbone=Qwen2.5-VL-7B2026.0640.18
PAPOGBackbone=Qwen2.5-VL-7B2026.0639.98
SFT-OracleBackbone=Qwen3-VL-2B-Instruct2026.0537.43
VISTA-SFTBackbone=Qwen3-VL-2B-Instruct2026.0536.27
R3VBackbone=Qwen3-VL-2B-Instruct2026.0535.77
STaRBackbone=Qwen3-VL-2B-Instruct2026.0534.28
STaR+Backbone=Qwen3-VL-2B-Instruct2026.0533.78
SFT-OracleBackbone=InternVL3-2B2026.0532.78
SFT-SeedBackbone=Qwen3-VL-2B-Instruct2026.0532.61
VISTA-SFTBackbone=InternVL3-2B2026.0531.45
STaR+Backbone=InternVL3-2B2026.0531.28
STaRBackbone=InternVL3-2B2026.0529.78
R3VBackbone=InternVL3-2B2026.0529.12
SFT-SeedBackbone=InternVL3-2B2026.0526.12