INDEX

Explanations

the word "English" specifically

oai_token-act-pair · gpt-3.5-turbo

references to the English language

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

Sensor

-0.71

ront

-0.68

 mounts

-0.68

orsi

-0.66

 incent

-0.66

@#

-0.66

[&

-0.64

 Romo

-0.64

raz

-0.63

kos

-0.63

POSITIVE LOGITS

 English

3.63

English

2.98

 english

2.84

english

1.93

 Spanish

1.88

 Arabic

1.84

 Portuguese

1.77

 Hindi

1.74

 French

1.72

 Welsh

1.69

Activations Density 0.022%

the word "English" specifically

references to the English language

No Comments

No Known Activations