List your company

Mathematical Reasoning on Countdown (Context Lengths 128, 256, 512)

54.3Accuracy (L=128)AIS
12.450423.315234.1845.0448Jan 12, 2026Feb 1, 2026Feb 21, 2026Mar 13, 2026Apr 2, 2026Apr 22, 2026May 13, 2026

Evaluation results

MethodLinks
AISBackbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.0554.357.0354.69—
BF16Backbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.0544.85252.2—
EvoToken (10k FT)NFE / Gen Len=12026.0139.8435.5542.9739.45
FlashRL (TIS)Backbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.0539.0642.9748.05—
FP8 RolloutBackbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.0537.7334.3833.2—
EvoToken (10k FT)NFE / Gen Len=1/22026.0134.7730.0830.0831.64
FT-Base (10k FT)NFE / Gen Len=12026.0133.221.4819.5324.74
FT-Base (10k FT)NFE / Gen Len=1/22026.0128.1216.816.4120.44
BaselineNFE / Gen Len=1/22026.0126.1716.4116.819.79
EvoToken (10k FT)NFE / Gen Len=1/42026.0123.0516.0212.1117.06
BaselineNFE / Gen Len=12026.0121.4823.8320.722
BaselineNFE / Gen Len=1/42026.0117.1915.6216.4116.41
FT-Base (10k FT)NFE / Gen Len=1/42026.0114.0613.679.7712.5