maximuspowers/muat-mean-std-pca-10-fourier-5-classifier
Updated • 1
Datasets:
example_id int64 0 10k | metadata stringlengths 679 723 | classification_prompt stringlengths 7.66k 18.5k | classification_completion stringclasses 14
values | classification_text stringlengths 7.68k 18.6k | improved_signature stringlengths 7.3k 18.8k | improved_model_weights stringlengths 1.77k 5.04k | training_metrics stringlengths 1.46k 2.92k |
|---|---|---|---|---|---|---|---|
0 | {"target_pattern": "sorted_descending", "degraded_accuracy": 0.7, "improved_accuracy": 0.94, "improvement": 0.24, "model_config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 5, "neurons_per_layer": 5, "activation_type": "gelu", "dropout_rate": 0.0, "random_seed": 9016, "learning_rate": 0.08961895813761998, "... | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 5
Neurons per Layer: 5
Activation Function: gelu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
0.723235,
0.148694,
-0.035862,
-0.03259,
... | sorted_descending | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 5
Neurons per Layer: 5
Activation Function: gelu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
0.723235,
0.148694,
-0.035862,
-0.03259,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"mean": 1.6626843214035034, "std": 1.5947465896606445, "pca": [-0.1394353061914444, 0.19910432398319244, 0.8945185542106628, 0.1676909327507019, 0.33560627698898315, 0.0, 0.0, 0.0, 0.0, 0.0], "fourier": [26.33926874630575, 27.472833786802177, 27.62489067195745, 32... | {"config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 5, "neurons_per_layer": 5, "activation_type": "gelu", "dropout_rate": 0.0, "precision": "float32", "input_size": 5, "input_format": "integer_indices"}, "weights": {"network.0.weight": [[0.723235, 0.148694, -0.035862, -0.03259, 0.093569], [0.391537, -0.74... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.6872678399085999, "train_acc": 0.565, "val_loss": 0.6929798722267151, "val_acc": 0.52}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.6848097145557404, "train_acc": 0.565, "val_loss": 0.6892448663711548, "v... |
1 | {"target_pattern": "palindrome", "degraded_accuracy": 0.48, "improved_accuracy": 0.94, "improvement": 0.45999999999999996, "model_config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 6, "neurons_per_layer": 7, "activation_type": "relu", "dropout_rate": 0.0, "random_seed": 2679, "learning_rate": 0.03008896643... | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 6
Neurons per Layer: 7
Activation Function: relu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
0.363811,
0.30837,
-0.057852,
-0.327621,
... | palindrome | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 6
Neurons per Layer: 7
Activation Function: relu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
0.363811,
0.30837,
-0.057852,
-0.327621,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"mean": 1.1651535034179688, "std": 1.6552081108093262, "pca": [0.47689467668533325, -0.10378384590148926, -0.16869188845157623, -0.33382144570350647, 0.5966607332229614, -0.07887818664312363, -0.5095889568328857, 0.0, 0.0, 0.0], "fourier": [26.69835519813661, 29.7... | {"config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 6, "neurons_per_layer": 7, "activation_type": "relu", "dropout_rate": 0.0, "precision": "float32", "input_size": 5, "input_format": "integer_indices"}, "weights": {"network.0.weight": [[0.363811, 0.30837, -0.057852, -0.327621, 0.637576], [0.398949, -0.37... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.678806722164154, "train_acc": 0.58, "val_loss": 0.7120968699455261, "val_acc": 0.48}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.678398996591568, "train_acc": 0.58, "val_loss": 0.7120597958564758, "val_a... |
2 | {"target_pattern": "alternating", "degraded_accuracy": 0.52, "improved_accuracy": 0.82, "improvement": 0.29999999999999993, "model_config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 5, "neurons_per_layer": 7, "activation_type": "gelu", "dropout_rate": 0.0, "random_seed": 9451, "learning_rate": 0.0735217037... | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 5
Neurons per Layer: 7
Activation Function: gelu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
-0.364313,
0.06905,
0.257017,
0.404946,
... | alternating | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 5
Neurons per Layer: 7
Activation Function: gelu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
-0.364313,
0.06905,
0.257017,
0.404946,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"mean": 2.0862905979156494, "std": 1.6003037691116333, "pca": [-0.3352741301059723, -0.12924033403396606, 0.6339672207832336, 0.05888080596923828, -0.19020946323871613, 0.38899242877960205, -0.5272685289382935, 0.0, 0.0, 0.0], "fourier": [24.40450067388025, 26.297... | {"config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 5, "neurons_per_layer": 7, "activation_type": "gelu", "dropout_rate": 0.0, "precision": "float32", "input_size": 5, "input_format": "integer_indices"}, "weights": {"network.0.weight": [[-0.364313, 0.06905, 0.257017, 0.404946, 0.532447], [-0.791896, -0.35... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.6916628181934357, "train_acc": 0.545, "val_loss": 0.6948902606964111, "val_acc": 0.52}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.656907469034195, "train_acc": 0.56, "val_loss": 0.6221851110458374, "val... |
3 | {"target_pattern": "increasing_pairs", "degraded_accuracy": 0.5, "improved_accuracy": 0.9, "improvement": 0.4, "model_config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 4, "neurons_per_layer": 6, "activation_type": "gelu", "dropout_rate": 0.0, "random_seed": 7902, "learning_rate": 0.019119242316001303, "ba... | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 4
Neurons per Layer: 6
Activation Function: gelu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
0.162133,
-0.334347,
0.141674,
-0.502033,
... | increasing_pairs | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 4
Neurons per Layer: 6
Activation Function: gelu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
0.162133,
-0.334347,
0.141674,
-0.502033,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"mean": -1.6593083143234253, "std": 1.4892473220825195, "pca": [-0.5073254704475403, 0.23020805418491364, 0.3740893602371216, 0.4705510437488556, 0.21680550277233124, -0.5303389430046082, 0.0, 0.0, 0.0, 0.0], "fourier": [20.438047015087, 21.858590541086098, 23.113... | {"config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 4, "neurons_per_layer": 6, "activation_type": "gelu", "dropout_rate": 0.0, "precision": "float32", "input_size": 5, "input_format": "integer_indices"}, "weights": {"network.0.weight": [[0.162133, -0.334347, 0.141674, -0.502033, -0.251815], [-0.204555, -0... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.7226835191249847, "train_acc": 0.425, "val_loss": 0.6928030252456665, "val_acc": 0.5}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.7003122270107269, "train_acc": 0.425, "val_loss": 0.6850212812423706, "va... |
4 | "{\"target_pattern\": \"starts_with\", \"degraded_accuracy\": 0.62, \"improved_accuracy\": 0.78, \"i(...TRUNCATED) | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | starts_with | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | "{\"neuron_activations\": {\"0\": {\"neuron_profiles\": {\"0\": {\"mean\": 1.2275820970535278, \"std(...TRUNCATED) | "{\"config\": {\"vocab_size\": 10, \"sequence_length\": 5, \"num_layers\": 4, \"neurons_per_layer\":(...TRUNCATED) | "{\"training_history\": [{\"stage\": \"degraded\", \"epoch\": 0, \"global_epoch\": 0, \"train_loss\"(...TRUNCATED) |
5 | "{\"target_pattern\": \"contains_abc\", \"degraded_accuracy\": 0.76, \"improved_accuracy\": 0.92, \"(...TRUNCATED) | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | contains_abc | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | "{\"neuron_activations\": {\"0\": {\"neuron_profiles\": {\"0\": {\"mean\": -2.566450834274292, \"std(...TRUNCATED) | "{\"config\": {\"vocab_size\": 10, \"sequence_length\": 5, \"num_layers\": 5, \"neurons_per_layer\":(...TRUNCATED) | "{\"training_history\": [{\"stage\": \"degraded\", \"epoch\": 0, \"global_epoch\": 0, \"train_loss\"(...TRUNCATED) |
6 | "{\"target_pattern\": \"increasing_pairs\", \"degraded_accuracy\": 0.64, \"improved_accuracy\": 0.88(...TRUNCATED) | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | increasing_pairs | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | "{\"neuron_activations\": {\"0\": {\"neuron_profiles\": {\"0\": {\"mean\": 1.2173339128494263, \"std(...TRUNCATED) | "{\"config\": {\"vocab_size\": 10, \"sequence_length\": 5, \"num_layers\": 4, \"neurons_per_layer\":(...TRUNCATED) | "{\"training_history\": [{\"stage\": \"degraded\", \"epoch\": 0, \"global_epoch\": 0, \"train_loss\"(...TRUNCATED) |
7 | "{\"target_pattern\": \"sorted_descending\", \"degraded_accuracy\": 0.56, \"improved_accuracy\": 0.9(...TRUNCATED) | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | sorted_descending | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | "{\"neuron_activations\": {\"0\": {\"neuron_profiles\": {\"0\": {\"mean\": -2.0215752124786377, \"st(...TRUNCATED) | "{\"config\": {\"vocab_size\": 10, \"sequence_length\": 5, \"num_layers\": 4, \"neurons_per_layer\":(...TRUNCATED) | "{\"training_history\": [{\"stage\": \"degraded\", \"epoch\": 0, \"global_epoch\": 0, \"train_loss\"(...TRUNCATED) |
8 | "{\"target_pattern\": \"has_majority\", \"degraded_accuracy\": 0.38, \"improved_accuracy\": 0.72, \"(...TRUNCATED) | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | has_majority | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | "{\"neuron_activations\": {\"0\": {\"neuron_profiles\": {\"0\": {\"mean\": -0.4140007495880127, \"st(...TRUNCATED) | "{\"config\": {\"vocab_size\": 10, \"sequence_length\": 5, \"num_layers\": 4, \"neurons_per_layer\":(...TRUNCATED) | "{\"training_history\": [{\"stage\": \"degraded\", \"epoch\": 0, \"global_epoch\": 0, \"train_loss\"(...TRUNCATED) |
9 | "{\"target_pattern\": \"decreasing_pairs\", \"degraded_accuracy\": 0.5, \"improved_accuracy\": 0.96,(...TRUNCATED) | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | decreasing_pairs | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | "{\"neuron_activations\": {\"0\": {\"neuron_profiles\": {\"0\": {\"mean\": 0.5709457993507385, \"std(...TRUNCATED) | "{\"config\": {\"vocab_size\": 10, \"sequence_length\": 5, \"num_layers\": 5, \"neurons_per_layer\":(...TRUNCATED) | "{\"training_history\": [{\"stage\": \"degraded\", \"epoch\": 0, \"global_epoch\": 0, \"train_loss\"(...TRUNCATED) |
End of preview. Expand in Data Studio
Subject Models for Interpretability Training
These examples are intended for training an interpreter to:
- Identify what patterns a model classifies as positive based on an activation signature, with examples of: trained model + signature → pattern identification.
| Signature Extraction | |
|---|---|
| Neuron Profile Methods | mean, std, pca, fourier |
| Prompt Format | separate |
| Signature Dataset | dataset_generation/exp_1/signature_dataset.json |
| Model Architecture | |
|---|---|
| Number of Layers | 4 to 6 |
| Neurons per Layer | 5 to 8 |
| Activation Types | relu, gelu |
| Pattern Vocab Size | 10 |
| Pattern Sequence Len | 5 |
| Training Datasets | |
|---|---|
| Enabled Patterns | palindrome, sorted_ascending, sorted_descending, alternating, contains_abc, starts_with, ends_with, no_repeats, has_majority, increasing_pairs, decreasing_pairs, vowel_consonant, first_last_match, mountain_pattern |
| Patterns per Batch | 1-1 |
| Pos/Neg Ratio | 1:1 |
| Target Total Examples per Subject Model | 250 |
| Staged Training | |
|---|---|
| Min Improvement Threshold | 0.05 (5.0%) |
| Corruption Rate | 0.15 (15.0%) |
Dataset Fields
| Field | Description |
|---|---|
| example_id | Unique identifier for each example |
| metadata | JSON string containing: |
- target_pattern: The pattern that was corrupted during training |
|
- degraded_accuracy: Accuracy of the model trained on corrupted data |
|
- improved_accuracy: Accuracy of the model after training on clean data |
|
- improvement: Delta between degraded and improved accuracy |
|
- model_config: Subject model architecture and hyperparameters |
|
- corruption_stats: Details about label corruption |
|
- selected_patterns: All patterns in the subject model's training dataset |
|
- precision: Model weight precision |
|
- quantization: Quantization type applied to weights |
|
- config_signature: Hash of critical config fields for validation |
|
| classification_prompt | Input prompt with improved model weights and signature |
| classification_completion | Target completion identifying the pattern |
| classification_text | Full concatenated text (prompt + completion) |
- Downloads last month
- 5
Models trained or fine-tuned on maximuspowers/muat-mean-std-pca-10-fourier-5