INDEX

Explanations

the word "pen" or variations of it

oai_token-act-pair · gpt-3.5-turbo

references to a "pen" or related contexts where the term is used

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

ONSORED

-0.76

 Dietary

-0.72

 Influence

-0.71

OHN

-0.70

ndum

-0.67

 Affect

-0.67

UTH

-0.66

 Giuliani

-0.65

Trend

-0.63

enance

-0.63

POSITIVE LOGITS

pen

1.39

ultimate

1.28

itent

1.21

 pens

1.16

elope

1.01

cil

0.99

manship

0.99

cill

0.98

insula

0.95

pen

0.94

Activations Density 0.006%

the word "pen" or variations of it

references to a "pen" or related contexts where the term is used

No Comments

No Known Activations