Commonsense Reasoning on Reasoning Suite Zero-shot Aggregate
73.2Aggregate ScoreLlama-2-13B
Evaluation Results
| Method | Links | |
|---|---|---|
| Llama-2-13B#Bit=16, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 73.2 | |
| ClusComp#Bit=4.09, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 73.1 | |
| ClusComp#Bit=2.81, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 72.2 | |
| GPTVQ#Bit=3.13, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 71.8 | |
| GPTQ#Bit=3.13, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 71.4 | |
| RTN#Bit=3.13, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 70.8 | |
| Llama-2-7B#Bit=16, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 70.5 | |
| ClusComp#Bit=4.14, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 69.6 | |
| ClusComp#Bit=2.19, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 69 | |
| ClusComp#Bit=2.14, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 68.4 | |
| ClusComp#Bit=2.89, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 67.9 | |
| GPTVQ#Bit=3.13, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 67.7 | |
| ClusComp#Bit=1.99, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 67.6 | |
| RTN#Bit=3.13, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 67.3 | |
| GPTQ#Bit=3.13, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 66.2 | |
| GPTVQ#Bit=2.25, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 66.2 | |
| FP16*#Bits=16-16, Evaluation Protocol=original paper result2026.05 | 66.01 | |
| Surrogate-Assisted Layer Contribution EstimationParameter Size=11.8B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 64.7 | |
| ShortGPTParameter Size=11.8B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 64.56 | |
| GPTVQ#Bit=2.13, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 64.5 | |
| ClusComp#Bit=2.29, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 64.1 | |
| SLEBParameter Size=11.8B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 63.93 | |
| FP16#Bits=16-162026.05 | 63.89 | |
| Shortened-LLaMAParameter Size=11.8B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 63.49 | |
| ClusComp#Bit=2.15, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 63 | |
| ClusComp#Bit=2, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 61.8 | |
| GPTVQ#Bit=2.25, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 61.5 | |
| Surrogate-Assisted Layer Contribution EstimationParameter Size=10.5B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 59.7 | |
| GEMQ#Bits=3-162026.05 | 59.49 | |
| ShortGPTParameter Size=10.5B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 58.78 | |
| SLEBParameter Size=10.5B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 57.62 | |
| MoEQuant*#Bits=3-16, Evaluation Protocol=original paper result2026.05 | 57.24 | |
| GPTVQ#Bit=2.13, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 57.2 | |
| GPTQ#Bit=2.25, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 54.2 | |
| GEMQ#Bits=2.5-162026.05 | 54.06 | |
| Shortened-LLaMAParameter Size=10.5B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 53.4 | |
| Surrogate-Assisted Layer Contribution EstimationParameter Size=9.2B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 53.27 | |
| SLEBParameter Size=9.2B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 52.89 | |
| SliceGPTParameter Size=11.8B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 49.59 | |
| Shortened-LLaMAParameter Size=9.2B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 48.25 | |
| ShortGPTParameter Size=9.2B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 48.25 | |
| GPTQ#Bit=2.25, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 47.6 | |
| GPTQ#Bit=2.13, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 46.6 | |
| RTN#Bit=2.25, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 46.4 | |
| SliceGPTParameter Size=10.5B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 42.86 | |
| RTN#Bit=2.25, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 42.4 | |
| RTN#Bit=2.13, Base Model=Llama-2-13B, Zero-shot=true2025.03 | 42.1 | |
| GPTQ#Bit=2.13, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 41.4 | |
| SliceGPTParameter Size=9.2B, Base Model=LLaMA-2-13B-hf, Evaluation Protocol=Zero-shot2026.02 | 39.5 | |
| RTN#Bit=2.13, Base Model=Llama-2-7B, Zero-shot=true2025.03 | 36.9 |