Classification Probing on Counterfact (test)
89.6Probe Acc (Best Layer)RAPTOR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RAPTORBackbone=Llama-3.1-70B-Instruct2026.01 | 89.6 | 82.2 | |
| RAPTORBackbone=Llama-3.1-8B-Instruct2026.01 | 89.3 | 80.3 | |
| xRFMBackbone=Llama-3.1-70B-Instruct2026.01 | 88.4 | 80 | |
| GCSBackbone=Llama-3.1-8B-Instruct2026.01 | 87.7 | 76 | |
| GCSBackbone=Llama-3.1-70B-Instruct2026.01 | 87.2 | 78.1 | |
| RAPTORBackbone=Llama-3.3-70B-Instruct2026.01 | 87.1 | 80 | |
| xRFMBackbone=Llama-3.1-8B-Instruct2026.01 | 86.9 | 78.8 | |
| RAPTORBackbone=Qwen-32B-Instruct2026.01 | 84.9 | 76.3 | |
| GCSBackbone=Llama-3.3-70B-Instruct2026.01 | 84.9 | 76 | |
| xRFMBackbone=Llama-3.3-70B-Instruct2026.01 | 84.5 | 77.3 | |
| xRFMBackbone=Qwen-32B-Instruct2026.01 | 83.5 | 74.9 | |
| GCSBackbone=Qwen-32B-Instruct2026.01 | 83.4 | 73.9 | |
| RAPTORBackbone=Qwen-7B-Instruct2026.01 | 82.3 | 72.3 | |
| RAPTORBackbone=Gemma-7B-it2026.01 | 80.9 | 71.2 | |
| RAPTORBackbone=Qwen-3B-Instruct2026.01 | 80.8 | 69.2 | |
| xRFMBackbone=Qwen-7B-Instruct2026.01 | 80.8 | 69.4 | |
| GCSBackbone=Qwen-7B-Instruct2026.01 | 79.8 | 69.4 | |
| xRFMBackbone=Gemma-7B-it2026.01 | 79.1 | 70.3 | |
| GCSBackbone=Gemma-7B-it2026.01 | 78.7 | 68.6 | |
| GCSBackbone=Qwen-3B-Instruct2026.01 | 78.5 | 65.2 | |
| xRFMBackbone=Qwen-3B-Instruct2026.01 | 77.6 | 67.2 |