INDEX

Explanations

proper nouns or names of companies, mostly preceded by news sources

oai_token-act-pair · gpt-3.5-turbo

instances of parentheses being used in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

rew

-0.74

ãĥł

-0.67

arton

-0.67

unia

-0.65

ishable

-0.63

itized

-0.63

ibles

-0.62

aunder

-0.62

hang

-0.62

 spam

-0.60

POSITIVE LOGITS

INESS

0.72

EVA

0.70

 Watkins

0.70

 Cohn

0.69

 Corps

0.68

 constitu

0.68

 Protective

0.66

 COUNTY

0.66

Thom

0.65

 corps

0.65

Activations Density 0.046%

proper nouns or names of companies, mostly preceded by news sources

instances of parentheses being used in the text

No Comments

No Known Activations