Classification on C. Chars
0.6129AccuracyUnsupervised Probe
Evaluation Results
| Method | Links | |
|---|---|---|
| Unsupervised ProbeModel=LLaMA3-8B2026.02 | 0.6129 | |
| Zero-Shot ProbeModel=LLaMA3-8B2026.02 | 0.5446 | |
| SAE ProbeModel=LLaMA3-8B2026.02 | 0.5428 | |
| SAE ProbeModel=Llama3.2-3B2026.02 | 0.5309 | |
| Unsupervised ProbeModel=Mistral2026.02 | 0.5233 | |
| SAE ProbeModel=Mistral2026.02 | 0.4995 | |
| Zero-Shot ProbeModel=Mistral2026.02 | 0.4716 | |
| Unsupervised ProbeModel=Llama3.2-3B2026.02 | 0.4569 | |
| Zero-Shot ProbeModel=Llama3.2-3B2026.02 | 0.4396 | |
| Text OutputModel=LLaMA3-8B2026.02 | 0.42 | |
| Text OutputModel=Llama3.2-3B2026.02 | 0.3333 | |
| Text OutputModel=Mistral2026.02 | 0.3333 | |
| Unsupervised ProbeModel=GPT2-Small2026.02 | 0.2814 | |
| Zero-Shot ProbeModel=GPT2-Small2026.02 | 0.2083 | |
| SAE ProbeModel=GPT2-Small2026.02 | 0.2021 | |
| Text OutputModel=GPT2-Small2026.02 | 0.17 |