diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..a9458aebcc1002c91b5c5dff565670596ea2ad4c
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.722100789313904,0.8148936033248901,0.7339815731820774,0.7362699198703521
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..a74b0512700c0e436e324cead891cd5078803b5a
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7258652094717669,0.8148936033248901,0.7339815731820774,0.7482856156778008
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..065f3ed9695b6aed48dd396d0ef77f5540a2b32b
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7233151183970856,0.8148936033248901,0.7339815731820774,0.7389334053599832
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..92fd19eb84a972af24f93dc4b8599e871ed5bbb6
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.725197328476017,0.8148936033248901,0.7339815731820774,0.7459710092734312
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..f8825182c07e89dab89d46dce5c2ab533bc6aab7
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7213114754098361,0.8148936033248901,0.7339815731820774,0.7376129167792084
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..ea470d634ca669877670585d2ec857d4f3b34a4d
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.7823922282938677,0.7989361882209778,0.7313669956812434,0.7467715004332847
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..79cdbd64311f65ff0f49e6de3c8cb87865c3883a
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.7822707953855494,0.7989361882209778,0.7313669956812434,0.746827862673402
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..edc0f5b9f8829261cb53fdbb977869f8beb5f4d9
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.7848208864602307,0.7989361882209778,0.7313669956812434,0.7369081084127688
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..63ec4c321471ed672498b461f5b9a9bf39d10a0a
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.7869763205828779,0.7989361882209778,0.7313669956812434,0.7408746010610191
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..ed635b16c1e93cf8d55128922866f786a2b8c68e
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_bce/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.7837279902853673,0.7989361882209778,0.7313669956812434,0.745623119790896
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..4d6cff976674d81d9f0fe5f122654ac2a301b75c
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7256223436551305,0.8148936033248901,0.7339815731820774,0.7424071906605443
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..bbb491875469f1b1de3bca0a8c13633f4db39841
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7248937462052216,0.8148936033248901,0.7339815731820774,0.7465674799675879
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..82ebda7317fa66599693f2ebb9b8be9b8b85a4b0
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7242865816636308,0.8148936033248901,0.7339815731820774,0.7460010203775397
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..279e5cbff9f8ea8d74006d496be319177766d1c7
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7238615664845173,0.8148936033248901,0.7339815731820774,0.7473552714504366
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..7e1fe049217119e05b0971540bd71535fbef42d5
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7232544019429266,0.8148936033248901,0.7339815731820774,0.7440277902824044
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..6b4a19a66c99d6f8a3f1609816fe6845bd272e6a
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.7824529447480266,0.7989361882209778,0.7313669956812434,0.7468771796335045
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..5f85742387ad671da53197d8207f502f33045e38
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.7823922282938677,0.7989361882209778,0.7313669956812434,0.7468208173933873
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..f9598993c7f86d69def3772995fff92b0007793e
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.784942319368549,0.7989361882209778,0.7313669956812434,0.7365276632919775
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..e2d25ea28a932b81ede955fc429a94adf92d6626
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.7863995142683667,0.7989361882209778,0.7313669956812434,0.7407055143406674
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..60c2559ef3a8341f0d1bbb3bb7023ab6811b955f
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_nol2sp/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.7834851244687311,0.7989361882209778,0.7313669956812434,0.7455738028307933
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..69b8b52c451d097867a1e91d2a2b08b98d0e3cd3
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7248330297510626,0.8148936033248901,0.7339815731820774,0.7370952252333363
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..2bc8a30ac7c575e94423e6825bf3266fc1fd98f1
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7240437158469946,0.8148936033248901,0.7339815731820774,0.7368101197443054
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..da6a49026cd23fb582d3d3352e17656e5232d6ef
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7246508803885853,0.8148936033248901,0.7339815731820774,0.7366525614477357
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..095b2a63175020c2bfc09b47a5c1874f1abf2ab8
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_noreg/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7219793564055859,0.8148936033248901,0.7339815731820774,0.7374853695867472
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..993676e3a05df1e60c53da74be4a0c1ab02e1302
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,ablation_entropy_finetune_cls_random_reg,100.0,0.1,1.0,cls_random,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7187006678809956,0.8148936033248901,0.7339815731820774,0.7094812580654843
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..2ef08db188d2433c0d005c51a8d2004d34a21077
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,ablation_entropy_finetune_cls_random_reg,100.0,0.1,1.0,cls_random,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.4721918639951426,0.8148936033248901,0.7339815731820774,0.4644218360793494
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..467d8de76c438a17b27768b55699cb39f90c9962
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,ablation_entropy_finetune_cls_random_reg,100.0,0.1,1.0,cls_random,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.5563448694596236,0.8148936033248901,0.7339815731820774,0.5091196242609766
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..b375fa5aedf2adfd9bcd0c6906b70eecfeec691b
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,ablation_entropy_finetune_cls_random_reg,100.0,0.1,1.0,cls_random,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7159077109896782,0.8148936033248901,0.7339815731820774,0.7294498964616908
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..24f4899d49b963500d8e9bb8860c7db9f60bc783
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,ablation_entropy_finetune_cls_random_reg,100.0,0.1,1.0,cls_random,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.7148148148148149,0.8148936033248901,0.7339815731820774,0.7218870982263437
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..7b4f551314220695cf8545e9c1765e71da9cc727
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,ablation_entropy_finetune_cls_random_reg,100.0,0.1,1.0,cls_random,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.3754705525197329,0.7989361882209778,0.7313669956812434,0.47526049922854186
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..57da7babdb4348cccc8d2c20400da61782fdbd34
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,ablation_entropy_finetune_cls_random_reg,100.0,0.1,1.0,cls_random,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.42185792349726775,0.7989361882209778,0.7313669956812434,0.4933316424661298
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..e3ab2f3d58d0cbdfb48ba0c15bba46dfcdee7674
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,ablation_entropy_finetune_cls_random_reg,100.0,0.1,1.0,cls_random,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.7924711596842744,0.7989361882209778,0.7313669956812434,0.7345338490478305
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..ec27a84f7075774a8d7a04d0827115210005b324
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,ablation_entropy_finetune_cls_random_reg,100.0,0.1,1.0,cls_random,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.4527018822100789,0.7989361882209778,0.7313669956812434,0.46256490464213507
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..9d27995575cc9240d0bff1ff9c4608015837adf2
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randcls/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,ablation_entropy_finetune_cls_random_reg,100.0,0.1,1.0,cls_random,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.42817243472981176,0.7989361882209778,0.7313669956812434,0.46514347712749843
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..e765ed77e10f4e1b66119a56cfb187719d5bb69e
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_0/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,ablation_entropy_finetune_linear_random_reg,100.0,0.1,1.0,linear_random,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.47128111718275645,0.8148936033248901,0.7339815731820774,0.5138351189940278
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..2efd0a8e1b8591c507490607fa8007eb60444ca0
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,ablation_entropy_finetune_linear_random_reg,100.0,0.1,1.0,linear_random,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.48803885853066176,0.8148936033248901,0.7339815731820774,0.5015605774136431
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..6210986102e943c63bd2cfd766196d07841ce9bb
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,ablation_entropy_finetune_linear_random_reg,100.0,0.1,1.0,linear_random,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.4611414693381906,0.8148936033248901,0.7339815731820774,0.47936736592539236
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..c007308776e63fdafc4cd7c1c20d4856c520c5ed
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,ablation_entropy_finetune_linear_random_reg,100.0,0.1,1.0,linear_random,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.4469945355191256,0.8148936033248901,0.7339815731820774,0.4770490081330092
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..d0d2c3d0efda64d5a3ded9b0395553d5d9e596b6
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,3,ablation_entropy_finetune_linear_random_reg,100.0,0.1,1.0,linear_random,0.0001,entropy,0,0.8495821952819824,0.7162720097146328,0.4831815421979356,0.8148936033248901,0.7339815731820774,0.45063173374148435
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..2cf134f2885918a93aac034a665905446826b596
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,ablation_entropy_finetune_linear_random_reg,100.0,0.1,1.0,linear_random,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.42756527018822105,0.7989361882209778,0.7313669956812434,0.43923798251361496
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..0ff924eda1fabc80a121a2bdf1618481bae4fe8c
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,ablation_entropy_finetune_linear_random_reg,100.0,0.1,1.0,linear_random,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.5258044930176077,0.7989361882209778,0.7313669956812434,0.556260083557021
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..5de1ebaac8434ffca13cd2596a3fbc0233270324
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,ablation_entropy_finetune_linear_random_reg,100.0,0.1,1.0,linear_random,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.45701275045537343,0.7989361882209778,0.7313669956812434,0.5198078047612003
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..f7d132a6747bf7afe84585701ae8ce80a5641387
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,ablation_entropy_finetune_linear_random_reg,100.0,0.1,1.0,linear_random,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.563873709775349,0.7989361882209778,0.7313669956812434,0.5778679573619653
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..b3477dbb148b6b0759ba24cb98279da926539b99
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_3/loo_randlin/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,3,ablation_entropy_finetune_linear_random_reg,100.0,0.1,1.0,linear_random,0.0001,entropy,0,0.8495821952819824,0.7785670916818459,0.6009714632665453,0.7989361882209778,0.7313669956812434,0.49002036085924233
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..c08477c495bd1e759c14e17ac56de9757ae47bbb
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.708133971291866,0.7946808338165283,0.7388552482815545,0.7480145105465038
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..e0c7e65ff362126f9e69e2a1effc6d53fc48519c
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7107655502392345,0.7946808338165283,0.7388552482815545,0.7488954089241249
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..a42697e64bb1852e0ba8044c1850ec17ba0eebc2
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7147129186602871,0.7946808338165283,0.7388552482815545,0.7452608360904759
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..79667f6acd8c0eecf350483d8ceee5d6bc2830c0
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr0.0001/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only,100.0,0.0,0.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7145933014354067,0.7946808338165283,0.7388552482815545,0.7465648431376629
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..ce2860272e056b2c957db89ed1c2d79ea98407c5
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only,100.0,0.0,0.0,cls_head,1e-05,entropy,0,0.8495821952819824,0.7448998178506375,0.7588949605343047,0.8085106611251831,0.7284393274853802,0.7496637426900583
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..3aee213d12f8b7574587bf828dbf8c2bbec87102
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only,100.0,0.0,0.0,cls_head,1e-05,entropy,0,0.8495821952819824,0.7448998178506375,0.757437765634487,0.8085106611251831,0.7284393274853802,0.7398391812865498
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..e76f62298f5f2b9c453a037583494155dee3a9e5
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only,100.0,0.0,0.0,cls_head,1e-05,entropy,0,0.8495821952819824,0.7448998178506375,0.7540376442015787,0.8085106611251831,0.7284393274853802,0.7452558479532163
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..038f33f5027dea9e6f3532a5fe0ddddf6dd36d79
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/bce_lr1e-05/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only,100.0,0.0,0.0,cls_head,1e-05,entropy,0,0.8495821952819824,0.7448998178506375,0.7570734669095324,0.8085106611251831,0.7284393274853802,0.7513230994152047
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_1/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_1/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
new file mode 100644
index 0000000000000000000000000000000000000000..f7ada1462fa914f4fe74346d422b7bfba2805194
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_1/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
@@ -0,0 +1,2 @@
+model,adapter,dataset,train_on_dataset,normalization,seed,test-acc,valid-confidnet-roc-auc,confidnet_roc_auc,best_lr,best_epoch,best_weighting,best_hidden,best_weight_decay
+roberta-base,LoraConfig,toxigen,train,True,4,0.7946808338165283,0.6786184210526316,0.7257423476288575,0.0003,30,1.0,400,0.0
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_2/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_2/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
new file mode 100644
index 0000000000000000000000000000000000000000..f2f1a1293dd63ac793828b991bf7b068ef20bf93
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_2/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
@@ -0,0 +1,2 @@
+model,adapter,dataset,train_on_dataset,normalization,seed,test-acc,valid-confidnet-roc-auc,confidnet_roc_auc,best_lr,best_epoch,best_weighting,best_hidden,best_weight_decay
+roberta-base,LoraConfig,toxigen,train,True,4,0.7946808338165283,0.7189593301435406,0.7260232640406185,0.0003,30,1.0,400,0.0
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_3/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_3/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
new file mode 100644
index 0000000000000000000000000000000000000000..527bba88730f226db73509c4fc1817a53521c96b
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_3/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
@@ -0,0 +1,2 @@
+model,adapter,dataset,train_on_dataset,normalization,seed,test-acc,valid-confidnet-roc-auc,confidnet_roc_auc,best_lr,best_epoch,best_weighting,best_hidden,best_weight_decay
+roberta-base,LoraConfig,toxigen,train,True,4,0.7946808338165283,0.6892643540669856,0.7272440365954318,0.0003,30,1.0,400,0.0
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_4/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_4/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
new file mode 100644
index 0000000000000000000000000000000000000000..b2a3c65f577b569d7fca12198ddca72d98a54fef
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep30_wd0.0/model_roberta/dataset_toxigen/head_seed_4/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
@@ -0,0 +1,2 @@
+model,adapter,dataset,train_on_dataset,normalization,seed,test-acc,valid-confidnet-roc-auc,confidnet_roc_auc,best_lr,best_epoch,best_weighting,best_hidden,best_weight_decay
+roberta-base,LoraConfig,toxigen,train,True,4,0.7946808338165283,0.6970992822966507,0.721521665244744,0.0003,30,1.0,400,0.0
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_1/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_1/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
new file mode 100644
index 0000000000000000000000000000000000000000..e2728031c10d0ab897d1b651631543d41f5ef818
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_1/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
@@ -0,0 +1,2 @@
+model,adapter,dataset,train_on_dataset,normalization,seed,test-acc,valid-confidnet-roc-auc,confidnet_roc_auc,best_lr,best_epoch,best_weighting,best_hidden,best_weight_decay
+google/electra-base-discriminator,LoraConfig,toxigen,train,True,4,0.8085106611251831,0.76381299332119,0.733954678362573,0.0003,50,1.0,400,0.0
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_2/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_2/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
new file mode 100644
index 0000000000000000000000000000000000000000..49659b6fc24a09de254e6a3b271f78855cf0b5c3
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_2/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
@@ -0,0 +1,2 @@
+model,adapter,dataset,train_on_dataset,normalization,seed,test-acc,valid-confidnet-roc-auc,confidnet_roc_auc,best_lr,best_epoch,best_weighting,best_hidden,best_weight_decay
+google/electra-base-discriminator,LoraConfig,toxigen,train,True,4,0.8085106611251831,0.7651487553126898,0.7319444444444444,0.0003,50,1.0,400,0.0
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_3/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_3/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
new file mode 100644
index 0000000000000000000000000000000000000000..7f5b814e6f92b754c2ca92fb46d0d5d9c456dbbd
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_3/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
@@ -0,0 +1,2 @@
+model,adapter,dataset,train_on_dataset,normalization,seed,test-acc,valid-confidnet-roc-auc,confidnet_roc_auc,best_lr,best_epoch,best_weighting,best_hidden,best_weight_decay
+google/electra-base-discriminator,LoraConfig,toxigen,train,True,4,0.8085106611251831,0.7625379477838494,0.7336915204678363,0.0003,50,1.0,400,0.0
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_4/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_4/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
new file mode 100644
index 0000000000000000000000000000000000000000..bca4e179df2ba08c3bb047af08b5d3f9e2e60e68
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/confidnet_lr0.0003_ep50_wd0.0/model_electra/dataset_toxigen/head_seed_4/LoraConfig_toxigen_seed_4_normalization_True_confidnet.csv
@@ -0,0 +1,2 @@
+model,adapter,dataset,train_on_dataset,normalization,seed,test-acc,valid-confidnet-roc-auc,confidnet_roc_auc,best_lr,best_epoch,best_weighting,best_hidden,best_weight_decay
+google/electra-base-discriminator,LoraConfig,toxigen,train,True,4,0.8085106611251831,0.7556162720097146,0.7372076023391813,0.0003,50,1.0,400,0.0
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..1507f0db42298b28a84dbae7d5419caed1d630c8
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune_l2sp_only,100.0,0.0,0.1,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.7656344869459624,0.8085106611251831,0.7284393274853802,0.7332748538011696
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..acf89aec1e8b0982b9f9c0440b56981c6ccca15a
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune_l2sp_only,100.0,0.0,0.1,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.7564663023679418,0.8085106611251831,0.7284393274853802,0.7350950292397661
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..dc0771ea3e1c03631dc0ccdc80f62288700e1595
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune_l2sp_only,100.0,0.0,0.1,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.7556769884638738,0.8085106611251831,0.7284393274853802,0.7460891812865498
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..fca138ca1af7c6b63548f127f7c2756e2fdbdd47
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr0.0001_b0.1/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune_l2sp_only,100.0,0.0,0.1,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.7618700667880997,0.8085106611251831,0.7284393274853802,0.7390570175438597
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..55ff3dcd6f882b1416c50ba81d98f8d9c1db0a4a
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune_l2sp_only,100.0,0.0,0.01,cls_head,1e-05,entropy,0,0.8467966318130493,0.6856758373205742,0.7033492822966508,0.7946808338165283,0.7388552482815545,0.7507265677563449
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..29ab88cdd389dbe3ecb9aaafc161502ad90bcfeb
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune_l2sp_only,100.0,0.0,0.01,cls_head,1e-05,entropy,0,0.8467966318130493,0.6856758373205742,0.7032894736842106,0.7946808338165283,0.7388552482815545,0.7506988229255537
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..11a60d600ba454e3e617ba72f8aafc0528599401
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune_l2sp_only,100.0,0.0,0.01,cls_head,1e-05,entropy,0,0.8467966318130493,0.6856758373205742,0.7058313397129187,0.7946808338165283,0.7388552482815545,0.7512953367875648
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..c84bd4ca4478ef18184994b400b33ed244bd17b0
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/l2sp_lr1e-05_b0.01/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune_l2sp_only,100.0,0.0,0.01,cls_head,1e-05,entropy,0,0.8467966318130493,0.6856758373205742,0.7050239234449761,0.7946808338165283,0.7388552482815545,0.7510803143489329
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..3c1b79793ea094a094be687c6cc1db0b930ba7ea
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.75591985428051,0.8085106611251831,0.7284393274853802,0.7464254385964912
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..c725e30ece9d25f16e2f619de2615a0ed4e64f8b
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.7577413479052824,0.8085106611251831,0.7284393274853802,0.7513742690058479
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..deebd5a7a6674dceff95bace597d95811ea515a5
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.7625379477838494,0.8085106611251831,0.7284393274853802,0.7443786549707602
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..5944537cfbf875296c361e1daa60cf7972fdca59
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_0/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7044258373205742,0.7946808338165283,0.7388552482815545,0.7507404401717405
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..3f13f14e8e39acbacc6d1c31b624072cf2dd3bb0
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7042464114832535,0.7946808338165283,0.7388552482815545,0.7508583557026031
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..fe7c96d689a4a612198517f838b6a2ac5a8222c1
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7035885167464114,0.7946808338165283,0.7388552482815545,0.7507404401717405
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..9b5fc90e318792288a1f00ec31a5f74321e293b9
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7036483253588517,0.7946808338165283,0.7388552482815545,0.7507612487948339
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..468dbbce5e538ef3b9e0ac3095d5629b62a5504c
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/loo_noreg/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune,100.0,0.0,1.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7038875598086124,0.7946808338165283,0.7388552482815545,0.7507126953409492
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..7053356f20a0c584cebba5b3d39c44686ad0ce79
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_cls_random,100.0,0.0,0.0,cls_random,0.0004,entropy,0,0.8495821952819824,0.7448998178506375,0.7431086824529448,0.8085106611251831,0.7284393274853802,0.7398611111111111
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..d45bf83180daf411c02b380eec94cce395da176c
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_cls_random,100.0,0.0,0.0,cls_random,0.0004,entropy,0,0.8495821952819824,0.7448998178506375,0.7482695810564663,0.8085106611251831,0.7284393274853802,0.7520285087719298
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..dffa73ba9db4bb2605033610d0862936b610b555
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_cls_random,100.0,0.0,0.0,cls_random,0.0004,entropy,0,0.8495821952819824,0.7448998178506375,0.7433515482695809,0.8085106611251831,0.7284393274853802,0.7347733918128656
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..783882c516a19cdccc29e3eef296bf4ae6a285d1
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_cls_random,100.0,0.0,0.0,cls_random,0.0004,entropy,0,0.8495821952819824,0.7448998178506375,0.7047965998785671,0.8085106611251831,0.7284393274853802,0.6789327485380117
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..b84ce8cbdc9288976b12e58fd1f2699493f81b96
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_cls_random,100.0,0.0,0.0,cls_random,0.0004,entropy,0,0.8467966318130493,0.6856758373205742,0.3556818181818182,0.7946808338165283,0.7388552482815545,0.42111104174903413
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..6a3db16c8dc137405ddc42d1f70db25fa6256492
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_cls_random,100.0,0.0,0.0,cls_random,0.0004,entropy,0,0.8467966318130493,0.6856758373205742,0.37607655502392345,0.7946808338165283,0.7388552482815545,0.41103273196412593
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..3aabc61db03df9f3bc2c80ecb267603d6ed23cba
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_cls_random,100.0,0.0,0.0,cls_random,0.0004,entropy,0,0.8467966318130493,0.6856758373205742,0.38648325358851676,0.7946808338165283,0.7388552482815545,0.4181978345159567
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..f28481039d8a14cdee8d46becfe10a57b55d8fac
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randcls_lr0.0004/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_cls_random,100.0,0.0,0.0,cls_random,0.0004,entropy,0,0.8467966318130493,0.6856758373205742,0.551854066985646,0.7946808338165283,0.7388552482815545,0.598560043281936
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..ce1d8b25a7790fd80f026667e5f9a17d777c6d5d
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_linear_random,100.0,0.0,0.0,linear_random,0.0004,entropy,0,0.8495821952819824,0.7448998178506375,0.7028536733454767,0.8085106611251831,0.7284393274853802,0.6579459064327485
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..cebf7936dff52eb323c9d70016529fc3d61efd13
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_linear_random,100.0,0.0,0.0,linear_random,0.0004,entropy,0,0.8495821952819824,0.7448998178506375,0.6854280510018215,0.8085106611251831,0.7284393274853802,0.6628143274853802
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..a411ea8edd218766b7677bc1917c5c0ad87c13dc
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_linear_random,100.0,0.0,0.0,linear_random,0.0004,entropy,0,0.8495821952819824,0.7448998178506375,0.7007285974499089,0.8085106611251831,0.7284393274853802,0.6510818713450294
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..4deade3a905c8bfe6bbb6145572769d3e2cebfc3
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_linear_random,100.0,0.0,0.0,linear_random,0.0004,entropy,0,0.8495821952819824,0.7448998178506375,0.6955069823922283,0.8085106611251831,0.7284393274853802,0.6390277777777778
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..32707b628d80b1c3fc4e7a57f054c79714e4ca25
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_linear_random,100.0,0.0,0.0,linear_random,0.0004,entropy,0,0.8467966318130493,0.6856758373205742,0.5495813397129186,0.7946808338165283,0.7388552482815545,0.5163382372321756
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..b006b3b4f26d9e14f96834bce5b8ae8c929f2d74
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_linear_random,100.0,0.0,0.0,linear_random,0.0004,entropy,0,0.8467966318130493,0.6856758373205742,0.3883373205741627,0.7946808338165283,0.7388552482815545,0.4249537008136171
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..71939568de10cc392bac117e4fbe824887de8448
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_linear_random,100.0,0.0,0.0,linear_random,0.0004,entropy,0,0.8467966318130493,0.6856758373205742,0.5413277511961723,0.7946808338165283,0.7388552482815545,0.5267078677403917
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..ac14458cd305047c9e1f3aea21033cd8b4cd7eff
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/randlin_lr0.0004/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,ablation_entropy_finetune_bce_only_linear_random,100.0,0.0,0.0,linear_random,0.0004,entropy,0,0.8467966318130493,0.6856758373205742,0.44748803827751193,0.7946808338165283,0.7388552482815545,0.4701777750032947
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..a4af7d6863147e67cf2a9f88ff9553152026fb6e
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune_reg_only,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.7546448087431694,0.8085106611251831,0.7284393274853802,0.7211330409356725
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..837d8fe98524dd4814d5586a5f3864d70a783f53
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune_reg_only,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.7591378263509411,0.8085106611251831,0.7284393274853802,0.7454459064327486
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..9caa9491752058bc2cbe1112f0fd2f1e6fd5295a
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune_reg_only,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.7554948391013965,0.8085106611251831,0.7284393274853802,0.7460526315789473
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..3e137bd2160a4bd6499cfe0c8c16c32957116ad3
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_electra/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,google/electra-base-discriminator,LoraConfig,toxigen,train,4,entropy_response_finetune_reg_only,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8495821952819824,0.7448998178506375,0.759805707346691,0.8085106611251831,0.7284393274853802,0.7169005847953217
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..d3f6b9e337cbaa85099e4dc2fcd1cf3fcdb26a49
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_1/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune_reg_only,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.708133971291866,0.7946808338165283,0.7388552482815545,0.7479867657157125
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..c615c1eb4c7f9ca9b373bf0f4890f6c258669289
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_2/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune_reg_only,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7107655502392345,0.7946808338165283,0.7388552482815545,0.7488884727164269
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..fc3b3b91aaa7c7721154fd497bec89830881a045
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_3/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune_reg_only,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7148325358851675,0.7946808338165283,0.7388552482815545,0.7452816447135694
diff --git a/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
new file mode 100644
index 0000000000000000000000000000000000000000..3e4641cbc7df5d1ca03c7d143ca839f0cff3125a
--- /dev/null
+++ b/revision_runs/heldout_v3/trainonly/runs/base_4/reg_lr0.0001_a0.1/model_roberta/dataset_toxigen/head_seed_4/best_run_metric_df.csv
@@ -0,0 +1,2 @@
+,model,adapter,dataset,train_on_dataset,seed,grid,lam,reg_alpha,l2sp_alpha,load_weights,lr,head_type,trial_number,valid-acc,valid-base-max-prob-roc-auc,valid-fine-tune-max-prob-roc-auc,test-acc,test-base-max-prob-roc-auc,test-fine-tune-max-prob-roc-auc
+0,roberta-base,LoraConfig,toxigen,train,4,entropy_response_finetune_reg_only,100.0,0.1,0.0,cls_head,0.0001,entropy,0,0.8467966318130493,0.6856758373205742,0.7145933014354067,0.7946808338165283,0.7388552482815545,0.747303549257479