INDEX

Explanations

the word "element" with a high activation value

oai_token-act-pair · gpt-3.5-turbo

the concept of "element" 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 @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

sburgh

-0.96

lishing

-0.72

pr

-0.69

blems

-0.67

blem

-0.65

claimed

-0.65

ulative

-0.65

ply

-0.63

trans

-0.63

ever

-0.63

POSITIVE LOGITS

 element

1.38

 elements

1.17

 Element

0.96

 "$:/

0.93

icide

0.83

idable

0.82

 abund

0.81

element

0.79

icides

0.74

osaurs

0.73

Activations Density 0.007%

the word "element" with a high activation value

the concept of "element" in various contexts

No Comments

No Known Activations