List your company

Math Word Problems on GSM8K (ACC and Length)

91.4AccuracyADaPTthink
46.57658.21369.8581.487Jun 18, 2026

Evaluation results

MethodLinks
ADaPTthinkBackbone=Qwen2.5-7B2026.0691.4803
SFT+GRPOBackbone=Qwen2.5-7B2026.0691.1801
ADaPTBackbone=Qwen2.5-7B2026.0690.8348
ARMBackbone=Qwen2.5-7B2026.0689.2297
SFT+GRPOBackbone=Qwen2.5-3B2026.0688.9873
R-4BBackbone=Qwen2.5-7B2026.0688.6310
ADaPTthinkBackbone=Qwen2.5-3B2026.0688.6816
TLMREBackbone=Qwen2.5-7B2026.0688.3584
ADaPTBackbone=Qwen2.5-3B2026.0686.3317
TLMREBackbone=Qwen2.5-3B2026.0685.8674
ADaPTanswerBackbone=Qwen2.5-7B2026.0684.7152
ARMBackbone=Qwen2.5-3B2026.0684.2286
R-4BBackbone=Qwen2.5-3B2026.0683.6427
BaseBackbone=Qwen2.5-7B2026.0681.8332
ADaPTanswerBackbone=Qwen2.5-3B2026.0680.1177
SFTBackbone=Qwen2.5-7B2026.0668.3541
BaseBackbone=Qwen2.5-3B2026.0664.1258
SFTBackbone=Qwen2.5-3B2026.0648.3194