File size: 274 Bytes
6a8ea55 | 1 2 3 4 5 6 7 | {
"interpretability_model": "Qwen/Qwen2.5-1.5B-Instruct",
"n_layers": 28,
"best_probe_layer": 23,
"best_probe_accuracy": 0.6514492753623189,
"note": "This model is a dedicated white-box interpretability testbed, distinct from the 16-model black-box leaderboard."
} |