List your company

Mathematical Reasoning on MATH (Pass@1, Pass@100)

46.2Pass@1LaDiR
4.0815.01525.9536.885Oct 6, 2025

Evaluation results

MethodLinks
LaDiRBackbone=LLAMA 3.1 8B2025.1046.263.7
TaH+Backbone=LLAMA 3.1 8B2025.1046.149.4
Soft TokenBackbone=LLAMA 3.1 8B2025.1044.7—
Soft ThinkBackbone=LLAMA 3.1 8B2025.1044.346.7
SFTBackbone=LLAMA 3.1 8B, Decoding temperature (alpha)=0.12025.1043.846.8
Discrete LatentBackbone=LLAMA 3.1 8B2025.1043.247.3
SFTBackbone=LLAMA 3.1 8B, Decoding temperature (alpha)=0.72025.1042.949.4
SFTBackbone=LLAMA 3.1 8B, Decoding temperature (alpha)=12025.1042.151
SFTBackbone=LLAMA 3.1 8B, Decoding temperature (alpha)=1.22025.1041.252.5
CoT SFTBackbone=LLADA 8B2025.103951
CODIBackbone=LLAMA 3.1 8B2025.1038.545.1
CoconutBackbone=LLAMA 3.1 8B2025.1037.339.3
Base ModelBackbone=LLADA 8B2025.1036.248.6
iCoTBackbone=LLAMA 3.1 8B2025.1035.237.9
LaDiR –w/o Stage 2Backbone=LLAMA 3.1 8B, Training Stage=without Stage 22025.1030.735.8
Pause TokenBackbone=LLAMA 3.1 8B2025.1027.231.4
Sol-Only SFTBackbone=LLAMA 3.1 8B2025.1013.318.9
LD4LGBackbone=LLAMA 3.1 8B2025.109.115.3
PLANNERBackbone=LLAMA 3.1 8B2025.105.713.1