List your company

Mathematical & Symbolic Reasoning on GSM8K (Tier 1)

89.2SRHalo
82.12883.96485.887.636Feb 22, 2026

Evaluation results

MethodLinks
HaloBackbone=LLaMA-3-70B2026.0289.21.29
CoT-ValveBackbone=LLaMA-3-70B2026.0288.90.85
AdaCoTBackbone=LLaMA-3-70B2026.0288.51.45
ToTBackbone=LLaMA-3-70B, b=5, d=32026.0287.43.5
CoT-SCBackbone=LLaMA-3-70B, k=102026.0286.12.15
Standard CoTBackbone=LLaMA-3-70B2026.0282.41