List your company

Mathematical Reasoning on GSM8K (100-question subset, test)

87AccuracyDALC-Id
30.8445.426074.58Dec 12, 2025Dec 30, 2025Jan 18, 2026Feb 6, 2026Feb 25, 2026Mar 16, 2026Apr 4, 2026

Evaluation results

MethodLinks
DALC-IdModel parameters=14B, n=1002026.04871,181
DALC-SVDModel parameters=14B, n=1002026.04861,200
SC (k=5)Model parameters=14B, n=1002026.04841,589
DALC-GSModel parameters=14B, n=1002026.04831,188
DALC-GSModel parameters=7B, n=1002026.04831,213
SnapKV-DBase Model=R1-Distill-Qwen14B, Cache budget=2562025.1282—
SeerAttentionBase Model=R1-Distill-Qwen14B, Cache budget=3842025.1282—
SingleModel parameters=14B, n=1002026.0482319
SC (k=5)Model parameters=7B, n=1002026.04821,648
DALC-SVDModel parameters=7B, n=1002026.04821,211
SnapKV-DBase Model=R1-Distill-Qwen14B, Cache budget=3842025.1281—
SnapKV-DBase Model=R1-Distill-Qwen14B, Cache budget=1282025.1280—
SeerAttentionBase Model=R1-Distill-Qwen14B, Cache budget=2562025.1280—
SnapKV-DBase Model=R1-Distill-Qwen14B, Cache budget=5122025.1278—
SingleModel parameters=7B, n=1002026.0477328
SeerAttentionBase Model=R1-Distill-Qwen14B, Cache budget=5122025.1270—
SeerAttentionBase Model=R1-Distill-Qwen14B, Cache budget=1282025.1266—
H2OBase Model=R1-Distill-Qwen14B, Cache budget=5122025.1264—
H2OBase Model=R1-Distill-Qwen14B, Cache budget=3842025.1262—
H2OBase Model=R1-Distill-Qwen14B, Cache budget=2562025.1256—
H2OBase Model=R1-Distill-Qwen14B, Cache budget=1282025.1233—