Local Causal Attribution on MATH 500
265.53AUPCAttriCoT
Evaluation results
| Method | Links | |
|---|---|---|
| AttriCoTModel=DS-Llama-8B2026.06 | 265.53 | |
| AttriCoT-2xModel=DS-Llama-8B2026.06 | 265.5 | |
| AttriCoTModel=DS-Qwen-14B2026.06 | 256.07 | |
| AttriCoT-2xModel=DS-Qwen-14B2026.06 | 255.83 | |
| TA-KL (no log)Model=DS-Llama-8B2026.06 | 234.93 | |
| TA-KL (no log)Model=DS-Qwen-14B2026.06 | 221.29 | |
| Prompted (best same-size)Model=DS-Llama-8B2026.06 | 181.33 | |
| Prompted-ICL (GPT-OSS)Model=DS-Llama-8B2026.06 | 179.94 | |
| Prompted-ICL (GPT-OSS)Model=DS-Qwen-14B2026.06 | 170.08 | |
| AttriCoT-2xModel=Qwen3-8B2026.06 | 148.4 | |
| AttriCoTModel=Qwen3-8B2026.06 | 147.56 | |
| Prompted (best same-size)Model=DS-Qwen-14B2026.06 | 143.61 | |
| TA-KL (no log)Model=Qwen3-8B2026.06 | 134.42 | |
| AttriCoT-2xModel=DS-Qwen3-8B2026.06 | 99.2 | |
| AttriCoTModel=DS-Qwen3-8B2026.06 | 98.38 | |
| TA-KL (no log)Model=DS-Qwen3-8B2026.06 | 84.16 | |
| Prompted-ICL (GPT-OSS)Model=Qwen3-8B2026.06 | 83.09 | |
| Prompted (best same-size)Model=Qwen3-8B2026.06 | 74.96 | |
| Prompted-ICL (GPT-OSS)Model=DS-Qwen3-8B2026.06 | 52.47 | |
| Prompted (best same-size)Model=DS-Qwen3-8B2026.06 | 44.86 |