ResearchBenchmarksMathematical & Symbolic Reasoning on MATH Tier 1Follow38.5Success Rate (SR)Halo9.69217.17124.6532.129Feb 22, 2026Evaluation ResultsMethodMethodLinksSuccess Rate (SR)Reasoning Attempts Count (RTO)HaloBackbone=LLaMA-3-70BBackbone=LLaMA-3-70B2026.0238.51.29CoT-ValveBackbone=LLaMA-3-70BBackbone=LLaMA-3-70B2026.0236.40.85AdaCoTBackbone=LLaMA-3-70BBackbone=LLaMA-3-70B2026.0219.31.45ToTBackbone=LLaMA-3-70B,...Backbone=LLaMA-3-70B, b=5, d=32026.0219.13.5CoT-SCBackbone=LLaMA-3-70B,...Backbone=LLaMA-3-70B, k=102026.0214.22.15Standard CoTBackbone=LLaMA-3-70BBackbone=LLaMA-3-70B2026.0210.81