List your company

Mathematical Reasoning on MultiArith (Acc, Avg Tok, Avg Steps)

43.33AccuracyCoT
19.649225.797131.94538.0929Mar 16, 2026

Evaluation results

MethodLinks
CoTModel=Llama-3.2-1B2026.0343.3328—
CoTModel=Qwen2.5-1.5B2026.0334.330.2—
AdaAnchor (adaptive)Model=Llama-3.2-1B2026.0332.442.573.5
iCoTModel=Llama-3.2-1B2026.0330.842.12—
AdaAnchor (adaptive)Model=Qwen2.5-1.5B2026.0329.42.163.82
iCoTModel=Qwen2.5-1.5B2026.0328.561.66—
AdaAnchor (K=8)Model=Llama-3.2-1B2026.0328.312.488
AdaAnchor (K=8)Model=Qwen2.5-1.5B2026.0327.62.348
No CoTModel=Qwen2.5-1.5B2026.0322.32.41—
No CoTModel=Llama-3.2-1B2026.0320.562.08—