List your company

Arithmetic Reasoning on SVAMP (Acc. & #L)

84.3AccuracyCoT-SFT
7.75627.62847.567.372Jun 29, 2026

Evaluation results

MethodLinks
CoT-SFTBackbone=Qwen3-VL-4B-Instruct2026.0684.355.9
CoT-SFTBackbone=Qwen3-4B2026.0682.953.7
RoTBackbone=Qwen3-VL-4B-Instruct2026.0672.732
DLR (Full)Backbone=Qwen3-VL-4B-Instruct2026.0672.73.8
DLR (Full)Backbone=Qwen3-4B2026.0672.33.8
DLR (SFT)Backbone=Qwen3-4B2026.0670.33.6
Direct-SFTBackbone=Qwen3-VL-4B-Instruct2026.06700
Direct-SFTBackbone=Qwen3-4B2026.0669.40
DLR (SFT)Backbone=Qwen3-VL-4B-Instruct2026.0668.73.6
CoLaRBackbone=Qwen3-4B2026.0657.719.2
CoLaRBackbone=Qwen3-VL-4B-Instruct2026.065218.4
CoconutBackbone=Qwen3-4B2026.0643.66
CoconutBackbone=Qwen3-VL-4B-Instruct2026.06426
iCoTBackbone=Qwen3-4B2026.0636.90
CODIBackbone=Qwen3-4B2026.06116
CODIBackbone=Qwen3-VL-4B-Instruct2026.0610.76