Zero-shot Evaluation on ARC, BoolQ, HellaSwag, LAMBADA, and MMLU benchmark leaderboard · SOTA2 Research

Zero-shot Evaluation on ARC, BoolQ, HellaSwag, LAMBADA, and MMLU

0.354ARC-CMuon-32

Evaluation Results

Method	Links
Muon-322025.09		0.354	0.713	0.618	0.428	0.424	0.261	0.466
Muon-8L/AdamW-322025.09		0.352	0.707	0.618	0.428	0.422	0.256	0.464
Muon-8L/AdamW-8D2025.09		0.347	0.706	0.625	0.425	0.418	0.26	0.463
Muon-8L/AdamW-322025.09		0.298	0.651	0.603	0.381	0.373	0.253	0.427
Muon-322025.09		0.296	0.653	0.607	0.382	0.376	0.255	0.428
Muon-8L/AdamW-8D2025.09		0.293	0.649	0.586	0.379	0.368	0.254	0.422
Muon-322025.09		0.281	0.62	0.606	0.355	0.342	0.25	0.409
Muon-8L/AdamW-322025.09		0.278	0.626	0.61	0.356	0.338	0.251	0.41
Muon-8L/AdamW-8D2025.09		0.272	0.628	0.611	0.355	0.335	0.245	0.408
Muon-8L/AdamW-322025.09		0.247	0.563	0.613	0.315	0.274	0.232	0.374
Muon-322025.09		0.236	0.564	0.606	0.316	0.272	0.231	0.371
Muon-8L/AdamW-8D2025.09		0.235	0.559	0.595	0.314	0.272	0.23	0.368
Muon-8L/AdamW-322025.09		0.211	0.505	0.574	0.285	0.202	0.23	0.335
Muon-8L/AdamW-8D2025.09		0.205	0.506	0.58	0.283	0.202	0.23	0.334
Muon-322025.09		0.204	0.503	0.571	0.285	0.203	0.23	0.332