List your company

Mathematical Reasoning on Spec-Bench Math

4.09CRSpecBound
1.34442.05722.773.4828Apr 14, 2026

Evaluation results

MethodLinks
SpecBoundBase Model=Vicuna-13B2026.044.091.91
SpecBoundBase Model=Vicuna-7B2026.043.781.89
SpecBoundBase Model=Codellama-7B-Instruct2026.043.632.02
SpecBoundBase Model=Codellama-13B-Instruct2026.043.492.2
KangarooBase Model=Vicuna-13B2026.042.421.63
AdaDecodeBase Model=Codellama-13B-Instruct2026.042.41.75
AdaDecodeBase Model=Codellama-7B-Instruct2026.042.21.62
SPACEBase Model=Vicuna-7B2026.042.151.65
KangarooBase Model=Vicuna-7B2026.042.141.61
LookaheadBase Model=Codellama-13B-Instruct2026.042.021.63
LookaheadBase Model=Vicuna-13B2026.041.981.61
LookaheadBase Model=Vicuna-7B2026.041.921.54
LookaheadBase Model=Codellama-7B-Instruct2026.041.881.52
MedusaBase Model=Codellama-7B-Instruct2026.041.781.96
MedusaBase Model=Vicuna-7B2026.041.771.96
MedusaBase Model=Vicuna-13B2026.041.762.06
MedusaBase Model=Codellama-13B-Instruct2026.041.742.04
RESTBase Model=Codellama-13B-Instruct2026.041.631.23
RESTBase Model=Vicuna-13B2026.041.591.21
RESTBase Model=Vicuna-7B2026.041.491.19
RESTBase Model=Codellama-7B-Instruct2026.041.451.17