Add files using upload-large-folder tool

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.12.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 2, "1": 43, "2": 194, "3": 653}, "net_accuracy": 0.29555997349237906}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.13.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 539, "1": 32, "2": 0, "3": 132}, "net_accuracy": 0.23293571901921803}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.13.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 645, "1": 0, "2": 0, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.14.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 2, "1": 0, "2": 12, "3": 864}, "net_accuracy": 0.29092113982770046}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.16.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 264, "1": 0, "2": 0, "3": 473}, "net_accuracy": 0.24420145791915177}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.16.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 645, "1": 0, "2": 0, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.17.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 0, "1": 718, "2": 0, "3": 1}, "net_accuracy": 0.23823724320742212}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.19.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 645, "1": 0, "2": 0, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.2.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 501, "1": 0, "2": 0, "3": 173}, "net_accuracy": 0.22332670642809807}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.2.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 645, "1": 0, "2": 0, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.20.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 124, "1": 167, "2": 328, "3": 114}, "net_accuracy": 0.24287607687210072}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.21.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 645, "1": 0, "2": 0, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.22.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 60, "1": 8, "2": 15, "3": 780}, "net_accuracy": 0.2859509609012591}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.23.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 15, "1": 61, "2": 604, "3": 129}, "net_accuracy": 0.2680583167660702}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.24.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 645, "1": 0, "2": 0, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.26.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 376, "1": 235, "2": 1, "3": 63}, "net_accuracy": 0.22365805168986083}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.27.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 132, "1": 506, "2": 83, "3": 9}, "net_accuracy": 0.24188204108681247}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.3.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 523, "1": 99, "2": 24, "3": 8}, "net_accuracy": 0.21669980119284293}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.3.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 645, "1": 0, "2": 0, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.32.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 644, "1": 0, "2": 1, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.33.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 640, "1": 1, "2": 5, "3": 0}, "net_accuracy": 0.2140490390987409}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.36.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 504, "1": 0, "2": 150, "3": 21}, "net_accuracy": 0.22365805168986083}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.37.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 550, "1": 0, "2": 99, "3": 6}, "net_accuracy": 0.2170311464546057}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.38.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 535, "1": 0, "2": 126, "3": 6}, "net_accuracy": 0.22100728959575877}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.39.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 510, "1": 0, "2": 89, "3": 78}, "net_accuracy": 0.22432074221338635}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.40.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 9, "1": 235, "2": 498, "3": 19}, "net_accuracy": 0.2521537442014579}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.40.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 498, "1": 0, "2": 125, "3": 40}, "net_accuracy": 0.21968190854870775}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.41.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 148, "1": 268, "2": 175, "3": 126}, "net_accuracy": 0.23757455268389663}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.41.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 512, "1": 0, "2": 134, "3": 21}, "net_accuracy": 0.22100728959575877}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.43.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 0, "1": 716, "2": 0, "3": 0}, "net_accuracy": 0.23724320742213387}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.44.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 624, "1": 4, "2": 8, "3": 11}, "net_accuracy": 0.21438038436050363}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.45.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 379, "1": 0, "2": 356, "3": 0}, "net_accuracy": 0.24353876739562624}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.45.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 499, "1": 0, "2": 172, "3": 0}, "net_accuracy": 0.22233267064280982}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.46.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 0, "1": 717, "2": 0, "3": 0}, "net_accuracy": 0.23757455268389663}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.5.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 585, "1": 0, "2": 2, "3": 48}, "net_accuracy": 0.21040424121935056}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.6.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 493, "1": 2, "2": 54, "3": 121}, "net_accuracy": 0.22200132538104705}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.6.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 645, "1": 0, "2": 0, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.7.hook_mlp_out.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 273, "1": 286, "2": 3, "3": 177}, "net_accuracy": 0.24486414844267726}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.7.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 645, "1": 0, "2": 0, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_blocks.8.hook_resid_post.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 645, "1": 0, "2": 0, "3": 0}, "net_accuracy": 0.21371769383697814}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_ln_final.hook_normalized.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 396, "1": 7, "2": 234, "3": 106}, "net_accuracy": 0.2461895294897283}

mmlu_STEM/few_shot_0/gpt2-xl/last_token_relative_depth_100/accuracy_output.json

@@ -0,0 +1 @@
+{"total_per_class": {"0": 645, "1": 719, "2": 783, "3": 871}, "correct_per_class": {"0": 396, "1": 7, "2": 234, "3": 106}, "net_accuracy": 0.2461895294897283}