List your company
ResearchFormerly WizWand

Mathematical Reasoning on GSM8k (Pass@1 Acc)

86Pass@1 Accuracy+10k SFT (Random)
42.5853.852565.12576.3975Apr 19, 2026
Updated 6mo ago

Evaluation results

MethodLinks
+10k SFT (Random)Backbone=Qwen2.5-3B, SFT examples=10k, Temperature=0.12026.0486
STAT-SynBackbone=Qwen2.5-3B, SFT examples=10k, Temperature=0.12026.0485
AUTOSKILLBackbone=Qwen2.5-3B, SFT examples=10k, Temperature=0.12026.0484.5
ReasonFlux-PRMBackbone=Qwen2.5-3B, SFT examples=10k, Temperature=0.12026.0481.5
AUTOSKILLBackbone=Llama3-8B-Instruct, SFT examples=1k, epochs=42026.0477.5
s1K-1.1 (DeepSeek)Backbone=Llama3-8B-Instruct, SFT examples=1k, epochs=42026.0477
+1k SFT (Random)Backbone=Llama3-8B-Instruct, SFT examples=1k, epochs=42026.0476.25
AUTOSKILLBackbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0473.43
Difficulty-lowBackbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0473
Remove-outliersBackbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0472.5
Long-examplesBackbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0472.25
Correctness-lowBackbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0472.25
DeduplicationBackbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0471.75
Repetition-lowBackbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0471.75
Best HeuristicBackbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0471.75
AUTOSKILLBackbone=Llama3-8B-Instruct, SFT examples=25k, epochs=12026.0470.75
Difficulty-highBackbone=Llama3-8B-Instruct, SFT examples=25k, epochs=12026.0470.62
Best HeuristicBackbone=Llama3-8B-Instruct, SFT examples=25k, epochs=12026.0470.62
Repetition-lowBackbone=Llama3-8B-Instruct, SFT examples=25k, epochs=12026.0470.56
+25k SFT (Random)Backbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0470.46
+25k SFT (Random)Backbone=Llama3-8B-Instruct, SFT examples=25k, epochs=12026.0470
Long-examplesBackbone=Llama3-8B-Instruct, SFT examples=25k, epochs=12026.0469
Correctness-highBackbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0468.5
LIMOBackbone=Llama3-8B-Instruct, SFT examples=1k, epochs=42026.0468.25
Difficulty-highBackbone=Qwen2.5-3B, SFT examples=25k, Temperature=1.02026.0467
ReasonFlux-PRMBackbone=Llama3-8B-Instruct, SFT examples=25k, epochs=12026.0465.75
Remove-outliersBackbone=Llama3-8B-Instruct, SFT examples=25k, epochs=12026.0465.62
Qwen2.5-3BBackbone=Qwen2.5-3B2026.0456.62
ReasonFlux-PRMBackbone=Llama3-8B-Instruct, SFT examples=1k, epochs=42026.0456
STAT-SelBackbone=Qwen2.5-3B, SFT examples=10k, Temperature=0.12026.0449
s1K-1.1 (Gemini)Backbone=Llama3-8B-Instruct, SFT examples=1k, epochs=42026.0444.75
Llama3-8B-InstructBackbone=Llama3-8B-Instruct2026.0444.25