List your company
ResearchFormerly WizWand

Logical Reasoning on BBH (test)

88.29Top@1 AccuracyFAA
41.833253.894165.95578.0159Dec 26, 2025Jan 14, 2026Feb 2, 2026Feb 21, 2026Mar 12, 2026Mar 31, 2026Apr 20, 2026
Updated 6mo ago

Evaluation results

MethodLinks
FAABackbone Model=Deepseek R1-1.5B, Trainable Parameters=1.63M2025.1288.29——
FFBackbone Model=Deepseek R1-1.5B, Trainable Parameters=1.5B2025.1288.27——
AdapterHBackbone Model=Deepseek R1-1.5B, Trainable Parameters=1.63M2025.1288.21——
LoRABackbone Model=Deepseek R1-1.5B, Trainable Parameters=1.21M2025.1288.17——
FourierFTBackbone Model=Deepseek R1-1.5B, Trainable Parameters=0.15M2025.1288.07——
Full ReasoningBase Model=DeepSeek-R1-Qwen2.5-14b2026.0185.2592.341,287
SyncThinkBase Model=DeepSeek-R1-Qwen2.5-14b2026.0183.8458.65817
T2TBackbone=LLaDA2.1-mini (16B MoE), Decoding Strategy=greedy decoding, Temperature=0, block length=322026.0475.5——
T2MBackbone=LLaDA2.1-mini (16B MoE), Decoding Strategy=greedy decoding, Temperature=0, block length=32, LOWPROB parameters=τ=0.3, Cmax=1, ρmax=0.252026.0475.38——
Full ReasoningBase Model=DeepSeek-R1-LLaMA3.1-8b2026.0173.3369.031,445
SyncThinkBase Model=DeepSeek-R1-LLaMA3.1-8b2026.0167.6838.24796
FAABackbone Model=Qwen2 7B, Trainable Parameters=7.30M2025.1267.1——
LoRABackbone Model=Qwen2 7B, Trainable Parameters=5.40M2025.1267.09——
FourierFTBackbone Model=Qwen2 7B, Trainable Parameters=0.85M2025.1267.05——
AdapterHBackbone Model=Qwen2 7B, Trainable Parameters=7.29M2025.1266.89——
FFBackbone Model=Qwen2 7B, Trainable Parameters=7.07B2025.1266.74——
Full ReasoningBase Model=DeepSeek-R1-Qwen2.5-7b2026.0165.8644.651,005
SyncThinkBase Model=DeepSeek-R1-Qwen2.5-7b2026.0163.0324.46541
No ReasoningBase Model=DeepSeek-R1-Qwen2.5-7b2026.0161.2116.08362
FAABackbone Model=LLaMA3 8B, Trainable Parameters=8.73M2025.1256.89——
FFBackbone Model=LLaMA3 8B, Trainable Parameters=8.03B2025.1256.87——
FourierFTBackbone Model=LLaMA3 8B, Trainable Parameters=0.91M2025.1256.85——
AdapterHBackbone Model=LLaMA3 8B, Trainable Parameters=8.73M2025.1256.82——
LoRABackbone Model=LLaMA3 8B, Trainable Parameters=6.47M2025.1256.81——
FAABackbone Model=LLaMA2 7B, Trainable Parameters=7.27M2025.1243.71——
LoRABackbone Model=LLaMA2 7B, Trainable Parameters=5.37M2025.1243.68——
FFBackbone Model=LLaMA2 7B, Trainable Parameters=6.94B2025.1243.67——
AdapterHBackbone Model=LLaMA2 7B, Trainable Parameters=7.27M2025.1243.65——
FourierFTBackbone Model=LLaMA2 7B, Trainable Parameters=0.82M2025.1243.62——