Context-aware generation on TriState-Bench Resistance corrupted context, prior correct (Sres)
33.25EMARR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ARRModel=Llama3-8B2026.06 | 33.25 | 38.59 | |
| ARRModel=Mistral-7B2026.06 | 22.5 | 30.66 | |
| ARRModel=Llama2-13B2026.06 | 17 | 23.21 | |
| ARRModel=Qwen2.5-7B2026.06 | 15.75 | 22.92 | |
| GreedyModel=Mistral-7B2026.06 | 5.25 | 12.99 | |
| GreedyModel=Llama3-8B2026.06 | 4.5 | 9.78 | |
| AdaCADModel=Llama3-8B2026.06 | 3.25 | 9.67 | |
| AdaCADModel=Mistral-7B2026.06 | 2.25 | 10.06 | |
| COIECDModel=Llama3-8B2026.06 | 1.75 | 6.29 | |
| GreedyModel=Llama2-13B2026.06 | 1.75 | 9.58 | |
| AdaCADModel=Llama2-13B2026.06 | 1.75 | 9.65 | |
| CoCoAModel=Llama2-13B2026.06 | 1.5 | 9.13 | |
| CoCoAModel=Llama3-8B2026.06 | 1 | 7.68 | |
| COIECDModel=Mistral-7B2026.06 | 1 | 8.61 | |
| GreedyModel=Qwen2.5-7B2026.06 | 1 | 10.04 | |
| COIECDModel=Llama2-13B2026.06 | 1 | 8.62 | |
| CADModel=Llama3-8B2026.06 | 0.75 | 7.11 | |
| AdaCADModel=Qwen2.5-7B2026.06 | 0.75 | 8.94 | |
| CoCoAModel=Mistral-7B2026.06 | 0.5 | 6.98 | |
| COIECDModel=Qwen2.5-7B2026.06 | 0.5 | 7.19 | |
| CADModel=Llama2-13B2026.06 | 0.5 | 7.75 | |
| CADModel=Mistral-7B2026.06 | 0.25 | 5.47 | |
| CADModel=Qwen2.5-7B2026.06 | 0.25 | 6.3 | |
| CoCoAModel=Qwen2.5-7B2026.06 | 0.25 | 6.19 |