General Language Evaluation on English lm-evaluation-harness benchmark leaderboard · SOTA2 Research

General Language Evaluation on English lm-evaluation-harness

0.819ARC Easy Acc (Norm)OjaKV

Evaluation Results

Method
OjaKV2025.09	0.819	—	0.5179	—	—	0.5914	—	—	—	0.7992	—	—	0.7395	0.6934
OjaKV2025.09	0.819	—	0.5188	—	—	0.5835	—	—	—	0.7982	—	—	0.7395	0.6918
Baseline2025.09	0.8186	—	0.5188	—	—	0.5904	—	—	—	0.8003	—	—	0.7388	0.6934
Eigen-N2025.09	0.8165	—	0.5154	—	—	0.5774	—	—	—	0.7965	—	—	0.7356	0.6883
StaticPCA2025.09	0.8165	—	0.5154	—	—	0.5774	—	—	—	0.7965	—	—	0.7356	0.6883
Eigen-N2025.09	0.7938	—	0.4812	—	—	0.5487	—	—	—	0.7867	—	—	0.6985	0.6618
StaticPCA2025.09	0.7938	—	0.4812	—	—	0.5487	—	—	—	0.7867	—	—	0.6985	0.6618
Baseline2025.09	0.7386	—	0.442	—	—	0.578	—	—	—	0.7639	—	—	0.6638	0.6373
OjaKV2025.09	0.7386	—	0.4437	—	—	0.5751	—	—	—	0.7579	—	—	0.663	0.6357
OjaKV2025.09	0.7374	—	0.4437	—	—	0.5706	—	—	—	0.7639	—	—	0.663	0.6357
Eigen-N2025.09	0.713	—	0.4138	—	—	0.5629	—	—	—	0.7503	—	—	0.659	0.6198
StaticPCA2025.09	0.713	—	0.4138	—	—	0.5629	—	—	—	0.7503	—	—	0.659	0.6198
Eigen-N2025.09	0.7003	—	0.3993	—	—	0.5499	—	—	—	0.7454	—	—	0.6298	0.6049
StaticPCA2025.09	0.7003	—	0.3993	—	—	0.5499	—	—	—	0.7454	—	—	0.6298	0.6049
Transformer + Spelling Bee Embeddings2026.01	0.5387	0.259	0.2764	0.0237	0.049	0.4247	0.2302	0.0169	0.31	0.6839	0.716	0.016	0.5375	—
Transformer2026.01	0.5173	0.2585	0.267	0.018	0.0382	0.4204	0.2331	0.0166	0.296	0.6773	0.676	0.0238	0.5295	—