List your company

Mathematical Reasoning on MATH500 (Accuracy and Solution Component Metrics)

91.2AccuracyFASA
3.42426.2124971.788Feb 3, 2026

Evaluation results

MethodLinks
FASABackbone=DeepSeek-R1-Distill-Qwen-14B, Fixed Budget=10002026.0291.21273,1393,266
FASABackbone=DeepSeek-R1-Distill-Qwen-32B, Fixed Budget=10002026.0291.21272,8873,014
FASABackbone=DeepSeek-R1-Distill-Qwen-14B, Fixed Budget=7002026.0290.21273,1393,266
FASABackbone=DeepSeek-R1-Distill-Qwen-32B, Fixed Budget=5002026.0290.21272,8873,014
FASABackbone=DeepSeek-R1-Distill-Qwen-32B, Fixed Budget=7002026.0290.21272,8873,014
FASABackbone=DeepSeek-R1-Distill-Qwen-14B, Fixed Budget=5002026.0288.81273,1393,266
FASABackbone=DeepSeek-R1-Distill-Qwen-14B, Fixed Budget=3002026.0286.61273,1393,266
FASABackbone=DeepSeek-R1-Distill-Qwen-32B, Fixed Budget=3002026.0286.41272,8873,014
OracleBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=7002026.0274.21273,1953,321
OracleBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=5002026.0272.61273,1953,321
FKVBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=3002026.0272.41272,9773,104
FKVBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=10002026.0272.41272,9773,104
OracleBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=10002026.0271.81273,1953,321
FASABackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=10002026.0271.81273,1713,298
OracleBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=3002026.0270.41273,1953,321
FASABackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=7002026.0269.41273,1713,298
FASABackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=5002026.0268.81273,1713,298
FASABackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=3002026.0262.21273,1713,298
RKVBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=10002026.02571277,0057,132
SnapKVBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=10002026.0254.61277,0477,174
H2OBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=7002026.0253.871278,2448,370
RKVBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=7002026.0249.21277,0057,132
StreamBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=10002026.0247.41273,5203,647
SnapKVBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=7002026.0246.81277,0477,174
H2OBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=10002026.0242.81278,2448,370
StreamBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=7002026.0240.41273,5203,647
RKVBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=5002026.0239.41277,0057,132
H2OBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=5002026.02331278,2448,370
SnapKVBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=5002026.0232.61277,0477,174
StreamBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=5002026.0224.61273,5203,647
RKVBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=3002026.02241277,0057,132
SnapKVBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=3002026.0221.61277,0477,174
StreamBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=3002026.029.61273,5203,647
H2OBackbone=DeepSeek-R1-Distill-Llama-8B, Fixed Budget=3002026.026.81278,2448,370