List your company

Mathematical Reasoning on GSM8K (Strict Accuracy, ALm, Rej)

93.1Accuracy (Strict)Cactus 1.0
83.407285.923688.4490.9564Apr 5, 2026

Evaluation results

MethodLinks
Cactus 1.0m (draft length)=10, Verifier model=Qwen 3 14B, Drafter model=Qwen 3 0.6B2026.0493.15.44-32
TASm (draft length)=20, Verifier model=Qwen 3 14B, Drafter model=Qwen 3 0.6B2026.0492.876.78-32
Cactus 1.0m (draft length)=20, Verifier model=Qwen 3 14B, Drafter model=Qwen 3 0.6B2026.0492.877-34
TASm (draft length)=10, Verifier model=Qwen 3 14B, Drafter model=Qwen 3 0.6B2026.0492.655.24-30
Cactus 0.75m (draft length)=20, Verifier model=Qwen 3 14B, Drafter model=Qwen 3 0.6B2026.0492.157.15-36
Cactus 0.75m (draft length)=10, Verifier model=Qwen 3 14B, Drafter model=Qwen 3 0.6B2026.0492.125.35-31
SpSm (draft length)=20, Verifier model=Qwen 3 14B, Drafter model=Qwen 3 0.6B2026.0491.895.11—
VerifierVerifier model=Qwen 3 14B2026.0491.71——
SpSm (draft length)=10, Verifier model=Qwen 3 14B, Drafter model=Qwen 3 0.6B2026.0491.124.27—
Cactus 0.75m (draft length)=20, Verifier model=Qwen 3 8B, Drafter model=Qwen 3 0.6B2026.0486.667.5-37
TASm (draft length)=10, Verifier model=Qwen 3 8B, Drafter model=Qwen 3 0.6B2026.0486.585.49-29
Cactus 1.0m (draft length)=20, Verifier model=Qwen 3 8B, Drafter model=Qwen 3 0.6B2026.0486.437.61-39
Cactus 1.0m (draft length)=10, Verifier model=Qwen 3 8B, Drafter model=Qwen 3 0.6B2026.0486.355.72-37
Cactus 0.75m (draft length)=10, Verifier model=Qwen 3 8B, Drafter model=Qwen 3 0.6B2026.0485.975.65-34
TASm (draft length)=20, Verifier model=Qwen 3 8B, Drafter model=Qwen 3 0.6B2026.0485.517.23-35
SpSm (draft length)=20, Verifier model=Qwen 3 8B, Drafter model=Qwen 3 0.6B2026.0484.465.44—
VerifierVerifier model=Qwen 3 8B2026.0484.31——
SpSm (draft length)=10, Verifier model=Qwen 3 8B, Drafter model=Qwen 3 0.6B2026.0483.784.49—