Explaining LLMs on SST
17.13CRRProxy Explanation Framework
Evaluation Results
| Method | Links | |
|---|---|---|
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 14B, Explanation Technique=LIME, CRR Type=local2025.05 | 17.13 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 14B, Explanation Technique=Kernel SHAP, CRR Type=local2025.05 | 17.13 | |
| Proxy Explanation FrameworkTarget Model=Llama 3.1 70B, Explanation Technique=Kernel SHAP, CRR Type=local2025.05 | 17.13 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 72B, Explanation Technique=LIME, CRR Type=local2025.05 | 16.25 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 72B, Explanation Technique=Kernel SHAP, CRR Type=local2025.05 | 16.25 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 32B, Explanation Technique=LIME, CRR Type=local2025.05 | 15.24 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 32B, Explanation Technique=Kernel SHAP, CRR Type=local2025.05 | 15.24 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o mini, Explanation Technique=Kernel SHAP, CRR Type=local2025.05 | 14.44 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o, Explanation Technique=LIME, CRR Type=local2025.05 | 14.17 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o, Explanation Technique=Kernel SHAP, CRR Type=local2025.05 | 14.17 | |
| Proxy Explanation FrameworkTarget Model=DeepSeek V3, Explanation Technique=Kernel SHAP, CRR Type=local2025.05 | 13.64 | |
| Proxy Explanation FrameworkTarget Model=DeepSeek V3, Explanation Technique=LIME, CRR Type=local2025.05 | 13.31 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o mini, Explanation Technique=LIME, CRR Type=local2025.05 | 13.15 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o, Explanation Technique=LIME, CRR Type=max2025.05 | 10.33 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o, Explanation Technique=Kernel SHAP, CRR Type=max2025.05 | 10.33 | |
| Proxy Explanation FrameworkTarget Model=Llama 3.1 70B, Explanation Technique=LIME, CRR Type=local2025.05 | 10.33 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o, Explanation Technique=Kernel SHAP, CRR Type=mean2025.05 | 8.9 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o, Explanation Technique=LIME, CRR Type=mean2025.05 | 8.74 | |
| Proxy Explanation FrameworkTarget Model=Llama 3.1 70B, Explanation Technique=Kernel SHAP, CRR Type=max2025.05 | 8.02 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 72B, Explanation Technique=LIME, CRR Type=max2025.05 | 7.04 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 72B, Explanation Technique=Kernel SHAP, CRR Type=max2025.05 | 6.91 | |
| Proxy Explanation FrameworkTarget Model=Llama 3.1 70B, Explanation Technique=LIME, CRR Type=max2025.05 | 6.74 | |
| Proxy Explanation FrameworkTarget Model=DeepSeek V3, Explanation Technique=Kernel SHAP, CRR Type=max2025.05 | 6.33 | |
| Proxy Explanation FrameworkTarget Model=Llama 3.1 70B, Explanation Technique=Kernel SHAP, CRR Type=mean2025.05 | 6.14 | |
| Proxy Explanation FrameworkTarget Model=Llama 3.1 70B, Explanation Technique=LIME, CRR Type=mean2025.05 | 5.32 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 72B, Explanation Technique=LIME, CRR Type=mean2025.05 | 5.1 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 72B, Explanation Technique=Kernel SHAP, CRR Type=mean2025.05 | 5.05 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 32B, Explanation Technique=LIME, CRR Type=max2025.05 | 4.77 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 32B, Explanation Technique=Kernel SHAP, CRR Type=max2025.05 | 4.77 | |
| Proxy Explanation FrameworkTarget Model=DeepSeek V3, Explanation Technique=LIME, CRR Type=max2025.05 | 4.6 | |
| Proxy Explanation FrameworkTarget Model=DeepSeek V3, Explanation Technique=Kernel SHAP, CRR Type=mean2025.05 | 3.88 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 32B, Explanation Technique=Kernel SHAP, CRR Type=mean2025.05 | 3.19 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 32B, Explanation Technique=LIME, CRR Type=mean2025.05 | 3.18 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o mini, Explanation Technique=Kernel SHAP, CRR Type=max2025.05 | 3.1 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o mini, Explanation Technique=LIME, CRR Type=max2025.05 | 3.08 | |
| Proxy Explanation FrameworkTarget Model=DeepSeek V3, Explanation Technique=LIME, CRR Type=mean2025.05 | 3.06 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 14B, Explanation Technique=LIME, CRR Type=max2025.05 | 2.9 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 14B, Explanation Technique=Kernel SHAP, CRR Type=max2025.05 | 2.89 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o mini, Explanation Technique=LIME, CRR Type=mean2025.05 | 2.5 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 14B, Explanation Technique=LIME, CRR Type=mean2025.05 | 2.39 | |
| Proxy Explanation FrameworkTarget Model=GPT-4o mini, Explanation Technique=Kernel SHAP, CRR Type=mean2025.05 | 1.97 | |
| Proxy Explanation FrameworkTarget Model=Qwen 2.5 14B, Explanation Technique=Kernel SHAP, CRR Type=mean2025.05 | 1.85 |