List your company

Mathematical Reasoning on GSM8K (Accuracy, Recall)

81AccuracyFull
26.50440.65254.868.948Nov 6, 2025Dec 15, 2025Jan 23, 2026Mar 4, 2026Apr 12, 2026May 21, 2026Jun 30, 2026

Evaluation results

MethodLinks
FullBackbone=LLaMA-3.1-8B-Instruct2026.0681100
RaBitQCacheBackbone=LLaMA-3.1-8B-Instruct2026.067788.8
DSBackbone=LLaMA-3.1-8B-Instruct2026.067772.7
SparQBackbone=LLaMA-3.1-8B-Instruct2026.067760.5
QuestBackbone=LLaMA-3.1-8B-Instruct2026.067081.3
BaselineSparsity=0%, Calib. data=-, Shot=8-shot2025.1157.6—
PuzzleMoESparsity=25%, Calib. data=C4, Shot=8-shot2025.1155.4—
PuzzleMoESparsity=50%, Calib. data=C4, Shot=8-shot2025.1151.7—
NAEESparsity=25%, Calib. data=Math, Shot=8-shot2025.1148.7—
NAEESparsity=25%, Calib. data=C4, Shot=8-shot2025.1141.5—
NAEESparsity=50%, Calib. data=Math, Shot=8-shot2025.1138.8—
WandaSparsity=50%, Calib. data=C4, Shot=8-shot2025.1132.2—
NAEESparsity=50%, Calib. data=C4, Shot=8-shot2025.1128.6—