List your company

Mathematical Reasoning on MATH 500 (Accuracy, Total Tokens)

83.6AccuracySC@5 + IRIS+
46.05655.80365.5575.297May 8, 2026

Evaluation results

MethodLinks
SC@5 + IRIS+n=500, Hardware=H800, Extraction Strategy=Fixed non-oracle agreement gate2026.0583.63,499
IRIS+Inference Budget (b)=4096, n=500, Hardware=H800, Extraction Strategy=Strengthened extraction2026.0578.82,645
SC@5 nothinkInference Budget (b)=1024, n=500, Hardware=H8002026.0576.62,685
IRISInference Budget (b)=4096, n=500, Hardware=H800, Extraction Strategy=full2026.05742,401
TOWN+Inference Budget (b)=4096, n=500, Hardware=H800, Extraction Strategy=Strengthened extraction2026.0572.42,565
TOWNInference Budget (b)=4096, n=500, Hardware=H800, Extraction Strategy=full2026.0571.82,565
ThinkInference Budget (b)=4096, n=500, Hardware=H8002026.05712,567
NothinkInference Budget (b)=4096, n=500, Hardware=H8002026.0568.6684
NothinkInference Budget (b)=2048, n=500, Hardware=H8002026.0568.4585
IRISInference Budget (b)=2048, n=500, Hardware=H800, Extraction Strategy=full2026.0567.21,573
NothinkInference Budget (b)=1024, n=500, Hardware=A1002026.0559.8606
SC@5 nothinkInference Budget (b)=512, n=500, Hardware=H8002026.05571,970
SC@3 nothinkInference Budget (b)=512, n=500, Hardware=H8002026.0555.81,182
TOWNInference Budget (b)=2048, n=500, Hardware=H800, Extraction Strategy=full2026.05551,590
ThinkInference Budget (b)=2048, n=500, Hardware=H8002026.0554.81,706
NothinkInference Budget (b)=512, n=200, Hardware=A100, Mode=pilot2026.0547.5418