List your company

Mathematics on GSM8K (held-out)

81.2AccuracyFixed Seed Image
23.01238.118553.22568.3315Jun 10, 2026

Evaluation results

MethodLinks
Fixed Seed ImageModel=Qwen3.5-2B2026.0681.2
Optimized Image (ART)Model=Qwen3.5-2B2026.0681.2
Random ImageModel=Qwen3.5-2B2026.0680.36
LoRA (DAPO)Model=Qwen3.5-2B2026.0677.33
BaselineModel=Qwen3.5-2B2026.0676.65
Random StringModel=Qwen3.5-2B2026.0671.49
Optimized Image (ART)Model=Qwen3.5-0.8B2026.0658.53
Fixed Seed ImageModel=Qwen3.5-0.8B2026.0656.33
Random ImageModel=Qwen3.5-0.8B2026.0654.59
LoRA (DAPO)Model=Qwen3.5-0.8B2026.0649.51
BaselineModel=Qwen3.5-0.8B2026.0639.65
Random StringModel=Qwen3.5-0.8B2026.0625.25