INDEX

Explanations

words related to trains and training

oai_token-act-pair · gpt-3.5-turbo

references to the word "train" in various contexts

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

uid

-0.75

hed

-0.74

hedral

-0.69

hern

-0.67

hedon

-0.66

 Fernandez

-0.66

 Gawker

-0.65

cens

-0.63

Wid

-0.62

arag

-0.62

POSITIVE LOGITS

 train

3.82

 trains

2.87

 Train

2.81

train

2.70

Train

2.55

 Amtrak

1.62

 railway

1.60

 training

1.55

 trained

1.54

 railroad

1.51

Activations Density 0.013%

words related to trains and training

references to the word "train" in various contexts

No Comments

No Known Activations