List your company

Mathematical Reasoning on MathBench

95.6AccuracyI-DLM
46.92859.56472.284.836Apr 13, 2026Apr 17, 2026Apr 21, 2026Apr 25, 2026Apr 29, 2026May 3, 2026May 7, 2026

Evaluation results

MethodLinks
I-DLMParams=32B, Decoding=ISD (N=4, sampling)2026.0495.6
Qwen3Params=32B2026.0495.5
Qwen3Params=8B2026.0493.1
I-DLMParams=8B, Decoding=ISD (N=4, sampling)2026.0489.1
LLaDA-2.1-miniParams=16B2026.0484.2
SDARParams=30B2026.0479.3
SDARParams=8B2026.0476.9
TaskFT-SPEED-24+BoSFine-tuning=Nemotron-Math, K=24, Visibility Policy=SPEED+BoS2026.0553
TaskFT-SPEED-20+BoSFine-tuning=Nemotron-Math, K=20, Visibility Policy=SPEED+BoS2026.0552.8
TaskFT-FullFine-tuning=Nemotron-Math, Depth=Full2026.0550.4
TaskFT-SPEED-28+BoSFine-tuning=Nemotron-Math, K=28, Visibility Policy=SPEED+BoS2026.0548.8