n_dataset_items: 300
n_items: 220
n_lie_condition_lied: 104
n_lie_condition_did_not_lie: 116
chosen_layer: 20
best_behavioral_layer: 31
top_k: 30
n_eval_features: 90
honest_lie_rate: 0
lie_cond_lie_rate: 0.4727272727
honest_mean_logit_diff: -14.53937766
lie_cond_mean_logit_diff: 2.28721147
lie_cond_lied_items_mean_logit_diff: 15.01502404
honest_accuracy: 1
behavioral_probe_test_auc_mean: 0.9852534562
behavioral_probe_test_auc_ci_low: 0.960506007
behavioral_probe_test_auc_ci_high: 0.9991046832
layer9_behavioral_probe_auc_mean: 0.6739170507
layer20_behavioral_probe_auc_mean: 0.9852534562
layer31_behavioral_probe_auc_mean: 0.9968663594
spearman_reliance_vs_causal: -0.01562332927
spearman_reliance_vs_binary_secondary: -0.1081249188
overlap_at_30: 0.1320754717
noncausal_correlate_frac: 1
random_feat_continuous_causal_p95: 0.1806640625
median_abs_continuous_causal_probe_top: 0.02313701923
median_abs_continuous_causal_random: 0.0478515625
random_feat_binary_delta_p95: 0
mean_honest_acc_drop_probe_top: 0.0001515151515
gpu_hours: 0.06488260408