List your company

Mathematical Reasoning on GSM8K (Correct, Success, JTax)

89AccuracyBenchmark LLM (BenchAcc)
56.7665.1373.581.87Jan 31, 2026

Evaluation results

MethodLinks
Benchmark LLM (BenchAcc)2026.0189——
Gradient Shiftvariant=hybrid2026.0188990
Gradient Shiftvariant=mult2026.01861003
Weak-to-Strongalpha=2.0, k=402026.0186981
Weak-to-Strongalpha=1.5, k=402026.0185971
Helper LLM2026.0185——
Weak-to-Strongalpha=1.0, k=402026.0181987
Gradient Shiftvariant=add2026.0158706