General Language Understanding on LLM Evaluation Suite (PiQA, ARC, HellaSwag, WinoGrande, MMLU v1)
74.6Overall AccuracyLLaMA-3-8B-Lizard
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLaMA-3-8B-LizardTraining Tokens (B)=0.04, Model Architecture Category=Linearized, Linearized Model Constraint=Unbounded2025.07 | 74.6 | 72.4 | |
| Mistral-7B-LoLCATsTraining Tokens (B)=0.04, Model Architecture Category=Linearized, Linearized Model Constraint=Bounded2025.07 | 74.5 | 70.7 | |
| Mistral-7B-LizardTraining Tokens (B)=0.04, Model Architecture Category=Linearized, Linearized Model Constraint=Unbounded2025.07 | 74.5 | 72.2 | |
| Mistral-7BTraining Tokens (B)=8000*, Model Architecture Category=Transformer2025.07 | 74.4 | 72.4 | |
| LLaMA-3-8B-LoLCATsTraining Tokens (B)=0.04, Model Architecture Category=Linearized, Linearized Model Constraint=Bounded2025.07 | 74.2 | 70.7 | |
| Gemma-7BTraining Tokens (B)=6000, Model Architecture Category=Transformer2025.07 | 74.1 | 72.3 | |
| LLaMA-3-8BTraining Tokens (B)=15000, Model Architecture Category=Transformer2025.07 | 73.1 | 72 | |
| Liger-GLA-Llama-3-8BTraining Tokens (B)=0.02, Model Architecture Category=Linearized, Linearized Model Constraint=Bounded2025.07 | 72.4 | 67.6 | |
| Griffin-7BTraining Tokens (B)=300, Model Architecture Category=Subquadratic2025.07 | 71.1 | 65.8 | |
| Mamba-7BTraining Tokens (B)=1200, Model Architecture Category=Subquadratic2025.07 | 71 | 64.7 | |
| Liger-GLA-Mistral-7BTraining Tokens (B)=0.02, Model Architecture Category=Linearized, Linearized Model Constraint=Bounded2025.07 | 70.9 | 65.1 | |
| Mistral-7B-SUPRATraining Tokens (B)=100, Model Architecture Category=Linearized, Linearized Model Constraint=Bounded2025.07 | 69.9 | 64 | |
| Hawk-7BTraining Tokens (B)=300, Model Architecture Category=Subquadratic2025.07 | 69.6 | 63.8 | |
| RWKV-6-v2.1-7BTraining Tokens (B)=1420, Model Architecture Category=Subquadratic2025.07 | 69.4 | 69.4 | |
| TransNormerLLM-7BTraining Tokens (B)=1400, Model Architecture Category=Subquadratic2025.07 | 68.2 | 64.1 | |
| Mamba2-LLaMA-3-8BTraining Tokens (B)=20, Model Architecture Category=Linearized, Linearized Model Constraint=Unbounded2025.07 | 65.6 | 61.9 |