Circuit Discovery on Previous-token
77ROC-AUCCo-activation
Evaluation Results
| Method | Links | |
|---|---|---|
| Co-activationModel=GPT-Neo-1.3B, Signal=co-activation2026.07 | 77 | |
| Conditional Co-ablationModel=GPT-2-sm, Signal=co-ablation2026.07 | 68 | |
| Co-activationModel=GPT-2-lg, Signal=co-activation2026.07 | 67 | |
| Conditional Co-ablationModel=Pythia-410M, Signal=co-ablation2026.07 | 66 | |
| Conditional Co-ablationModel=GPT-2-md, Signal=co-ablation2026.07 | 64 | |
| Conditional Co-ablationModel=Pythia-1.4B, Signal=co-ablation2026.07 | 64 | |
| Conditional Co-ablationModel=GPT-2-lg, Signal=co-ablation2026.07 | 60 | |
| Conditional Co-ablationModel=Gemma-2-2B, Signal=co-ablation2026.07 | 60 | |
| Conditional Co-ablationModel=Pythia-160M, Signal=co-ablation2026.07 | 56 | |
| Conditional Co-ablationModel=OLMo-2-7B, Signal=co-ablation2026.07 | 56 | |
| Conditional Co-ablationModel=Llama-3.1-8B, Signal=co-ablation2026.07 | 54 | |
| Co-activationModel=GPT-2-sm, Signal=co-activation2026.07 | 52 | |
| Co-activationModel=Gemma-2-2B, Signal=co-activation2026.07 | 46 | |
| Conditional Co-ablationModel=GPT-Neo-1.3B, Signal=co-ablation2026.07 | 45 | |
| Conditional Co-ablationModel=Gemma-2-9B, Signal=co-ablation2026.07 | 43 | |
| Co-activationModel=Pythia-160M, Signal=co-activation2026.07 | 42 | |
| Conditional Co-ablationModel=Qwen2.5-7B, Signal=co-ablation2026.07 | 41 | |
| Co-activationModel=Gemma-2-9B, Signal=co-activation2026.07 | 40 | |
| Co-activationModel=GPT-2-md, Signal=co-activation2026.07 | 39 | |
| Co-activationModel=Pythia-410M, Signal=co-activation2026.07 | 36 | |
| Co-activationModel=OLMo-2-7B, Signal=co-activation2026.07 | 36 | |
| Co-activationModel=Llama-3.1-8B, Signal=co-activation2026.07 | 31 | |
| Co-activationModel=Pythia-1.4B, Signal=co-activation2026.07 | 28 | |
| Co-activationModel=Qwen2.5-7B, Signal=co-activation2026.07 | 24 |