maximuspowers/muat-fourier-3-classifier
Updated • 1
Datasets:
example_id int64 0 10k | metadata stringlengths 679 723 | classification_prompt stringlengths 4.51k 10.9k | classification_completion stringclasses 14
values | classification_text stringlengths 4.52k 11k | improved_signature stringlengths 2.44k 4.91k | improved_model_weights stringlengths 1.76k 5.03k | training_metrics stringlengths 1.46k 2.92k |
|---|---|---|---|---|---|---|---|
0 | {"target_pattern": "sorted_descending", "degraded_accuracy": 0.4, "improved_accuracy": 0.96, "improvement": 0.5599999999999999, "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": 0.089618... | ## 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": [
[
-0.135936,
-0.898229,
0.440865,
0.123276,
... | 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": [
[
-0.135936,
-0.898229,
0.440865,
0.123276,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"fourier": [34.38320379607026, 40.171638428931615, 40.35054685021329]}, "1": {"fourier": [22.239334893166323, 23.422165914417118, 36.439215302467346]}, "2": {"fourier": [24.708134034066976, 25.798288262531763, 114.1671646386385]}, "3": {"fourier": [20.733267764821... | {"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": [[-0.135936, -0.898229, 0.440865, 0.123276, 0.184175], [-0.163952, -0.... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.6779561638832092, "train_acc": 0.445, "val_loss": 0.6844655871391296, "val_acc": 0.4}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.5416218340396881, "train_acc": 0.59, "val_loss": 0.5125601887702942, "val... |
1 | {"target_pattern": "palindrome", "degraded_accuracy": 0.54, "improved_accuracy": 0.92, "improvement": 0.38, "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": 0.03008896643339405, "batch_... | ## 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": [
[
0.125469,
-0.22807,
0.229685,
-0.28477,
... | 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": [
[
0.125469,
-0.22807,
0.229685,
-0.28477,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"fourier": [14.851885406020168, 15.327329002609316, 15.898824652225157]}, "1": {"fourier": [22.96374293503703, 24.609837488875996, 27.6968475576902]}, "2": {"fourier": [30.389728259886475, 33.531172914776825, 185.28660035133362]}, "3": {"fourier": [32.807046299186... | {"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": [[0.125469, -0.22807, 0.229685, -0.28477, -0.072223], [0.568673, 0.057... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.6893084645271301, "train_acc": 0.565, "val_loss": 0.6865367293357849, "val_acc": 0.54}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.6810666918754578, "train_acc": 0.565, "val_loss": 0.6784811615943909, "v... |
2 | {"target_pattern": "alternating", "degraded_accuracy": 0.52, "improved_accuracy": 0.86, "improvement": 0.33999999999999997, "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": [
[
-0.551271,
-0.05203,
0.138717,
0.324818,
... | 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": [
[
-0.551271,
-0.05203,
0.138717,
0.324818,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"fourier": [25.244357064938395, 25.272165219150512, 69.59754720330238]}, "1": {"fourier": [48.9682509831439, 51.870677516907044, 261.766864720732]}, "2": {"fourier": [30.022897412038205, 32.90533997602122, 32.93529503047466]}, "3": {"fourier": [48.42734209241126, ... | {"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": [[-0.551271, -0.05203, 0.138717, 0.324818, 0.35961], [-1.061444, -0.50... | {"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": [
[
0.308248,
-0.106554,
0.317253,
-0.270284,
... | 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": [
[
0.308248,
-0.106554,
0.317253,
-0.270284,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"fourier": [19.862496341990465, 20.830531205643247, 23.455964750442998]}, "1": {"fourier": [29.712725321245426, 31.584885187962293, 209.921728387475]}, "2": {"fourier": [19.809703701121816, 24.477207544049705, 56.49178962409496]}, "3": {"fourier": [24.325519491898... | {"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": [[0.308248, -0.106554, 0.317253, -0.270284, -0.076717], [-0.130905, 0.... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.7180013060569763, "train_acc": 0.425, "val_loss": 0.6934688091278076, "val_acc": 0.5}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.7019679844379425, "train_acc": 0.425, "val_loss": 0.6853039264678955, "va... |
4 | {"target_pattern": "increasing_pairs", "degraded_accuracy": 0.64, "improved_accuracy": 0.86, "improvement": 0.21999999999999997, "model_config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 4, "neurons_per_layer": 5, "activation_type": "relu", "dropout_rate": 0.0, "random_seed": 9859, "learning_rate": 0.01538... | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 4
Neurons per Layer: 5
Activation Function: relu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
0.490115,
-0.014935,
0.309498,
0.129559,
... | increasing_pairs | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 4
Neurons per Layer: 5
Activation Function: relu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
0.490115,
-0.014935,
0.309498,
0.129559,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"fourier": [25.760289924081444, 26.184900339264573, 86.41287272423506]}, "1": {"fourier": [24.9155429796375, 26.225291385790893, 178.7224967032671]}, "2": {"fourier": [29.024818433880576, 32.599501199529016, 168.50399085879326]}, "3": {"fourier": [26.1890723478261... | {"config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 4, "neurons_per_layer": 5, "activation_type": "relu", "dropout_rate": 0.0, "precision": "float32", "input_size": 5, "input_format": "integer_indices"}, "weights": {"network.0.weight": [[0.490115, -0.014935, 0.309498, 0.129559, -0.346239], [-0.099157, -0.... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.69880810379982, "train_acc": 0.455, "val_loss": 0.700002908706665, "val_acc": 0.36}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.6939691603183746, "train_acc": 0.465, "val_loss": 0.6915084719657898, "val_... |
5 | {"target_pattern": "sorted_descending", "degraded_accuracy": 0.56, "improved_accuracy": 0.94, "improvement": 0.3799999999999999, "model_config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 4, "neurons_per_layer": 8, "activation_type": "relu", "dropout_rate": 0.0, "random_seed": 4854, "learning_rate": 0.09414... | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 4
Neurons per Layer: 8
Activation Function: relu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
-0.001613,
-0.651691,
0.079222,
-0.03881,
... | sorted_descending | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 4
Neurons per Layer: 8
Activation Function: relu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
-0.001613,
-0.651691,
0.079222,
-0.03881,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"fourier": [22.509394282314062, 25.808451994717814, 181.94177502393723]}, "1": {"fourier": [33.62448678864054, 34.69252648210499, 145.4127191901207]}, "2": {"fourier": [24.11078469696945, 24.947207663878988, 96.0818722397089]}, "3": {"fourier": [34.78544813435153,... | {"config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 4, "neurons_per_layer": 8, "activation_type": "relu", "dropout_rate": 0.0, "precision": "float32", "input_size": 5, "input_format": "integer_indices"}, "weights": {"network.0.weight": [[-0.001613, -0.651691, 0.079222, -0.03881, -0.174839], [-0.546113, -0... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.6972275674343109, "train_acc": 0.565, "val_loss": 0.6631535291671753, "val_acc": 0.56}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.640335738658905, "train_acc": 0.565, "val_loss": 0.5430698394775391, "va... |
6 | {"target_pattern": "has_majority", "degraded_accuracy": 0.38, "improved_accuracy": 0.76, "improvement": 0.38, "model_config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 4, "neurons_per_layer": 5, "activation_type": "gelu", "dropout_rate": 0.0, "random_seed": 8556, "learning_rate": 0.09363094593146719, "batc... | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 4
Neurons per Layer: 5
Activation Function: gelu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
-0.111184,
0.079379,
0.119502,
0.467865,
... | has_majority | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 4
Neurons per Layer: 5
Activation Function: gelu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
-0.111184,
0.079379,
0.119502,
0.467865,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"fourier": [19.231210870143794, 20.091906388056312, 121.3213138282299]}, "1": {"fourier": [20.62894000032097, 23.636099345782963, 120.72935879230499]}, "2": {"fourier": [29.02673154117011, 31.055191413076425, 31.20639655457191]}, "3": {"fourier": [58.9849568181424... | {"config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 4, "neurons_per_layer": 5, "activation_type": "gelu", "dropout_rate": 0.0, "precision": "float32", "input_size": 5, "input_format": "integer_indices"}, "weights": {"network.0.weight": [[-0.111184, 0.079379, 0.119502, 0.467865, 0.150234], [-0.149945, 0.08... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.6833219230175018, "train_acc": 0.6, "val_loss": 0.760266900062561, "val_acc": 0.38}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.6676255762577057, "train_acc": 0.6, "val_loss": 0.772173285484314, "val_acc... |
7 | {"target_pattern": "decreasing_pairs", "degraded_accuracy": 0.5, "improved_accuracy": 0.98, "improvement": 0.48, "model_config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 5, "neurons_per_layer": 8, "activation_type": "relu", "dropout_rate": 0.0, "random_seed": 9319, "learning_rate": 0.04720846293947128, "b... | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 5
Neurons per Layer: 8
Activation Function: relu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
-0.734558,
0.403281,
0.109562,
-0.157404,
... | decreasing_pairs | ## Model Architecture
Input Size: 5 (integer indices for 5 sequence positions, vocab size 10)
Hidden Layers: 5
Neurons per Layer: 8
Activation Function: relu
Dropout Rate: 0.0
## Model Weights
The trained model weights:
{
"network.0.weight": [
[
-0.734558,
0.403281,
0.109562,
-0.157404,
... | {"neuron_activations": {"0": {"neuron_profiles": {"0": {"fourier": [24.239907579149094, 25.985161768157088, 30.55240732436437]}, "1": {"fourier": [40.862095123672475, 45.993506645281236, 189.91553899645805]}, "2": {"fourier": [27.677036194663874, 28.506403050687098, 203.1213087104261]}, "3": {"fourier": [29.03411603408... | {"config": {"vocab_size": 10, "sequence_length": 5, "num_layers": 5, "neurons_per_layer": 8, "activation_type": "relu", "dropout_rate": 0.0, "precision": "float32", "input_size": 5, "input_format": "integer_indices"}, "weights": {"network.0.weight": [[-0.734558, 0.403281, 0.109562, -0.157404, 0.124721], [0.772002, 0.62... | {"training_history": [{"stage": "degraded", "epoch": 0, "global_epoch": 0, "train_loss": 0.6809704303741455, "train_acc": 0.57, "val_loss": 0.7026543617248535, "val_acc": 0.5}, {"stage": "degraded", "epoch": 1, "global_epoch": 1, "train_loss": 0.6687776148319244, "train_acc": 0.57, "val_loss": 0.6631693840026855, "val_... |
8 | "{\"target_pattern\": \"decreasing_pairs\", \"degraded_accuracy\": 0.42, \"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\": {\"fourier\": [39.35651230091677, 41(...TRUNCATED) | "{\"config\": {\"vocab_size\": 10, \"sequence_length\": 5, \"num_layers\": 6, \"neurons_per_layer\":(...TRUNCATED) | "{\"training_history\": [{\"stage\": \"degraded\", \"epoch\": 0, \"global_epoch\": 0, \"train_loss\"(...TRUNCATED) |
9 | "{\"target_pattern\": \"first_last_match\", \"degraded_accuracy\": 0.64, \"improved_accuracy\": 0.86(...TRUNCATED) | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | first_last_match | "## Model Architecture\nInput Size: 5 (integer indices for 5 sequence positions, vocab size 10)\nHid(...TRUNCATED) | "{\"neuron_activations\": {\"0\": {\"neuron_profiles\": {\"0\": {\"fourier\": [37.43054472797677, 44(...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) |
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 | 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
- 10
Models trained or fine-tuned on maximuspowers/muat-fourier-3