INDEX

Explanations

high activation values for the word "draw."

oai_token-act-pair · gpt-3.5-turbo

instances of the word "draw" and its variations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

 ILCS

-0.81

ican

-0.79

nown

-0.78

ailability

-0.73

xon

-0.71

illon

-0.70

bda

-0.70

icum

-0.70

onte

-0.67

berra

-0.64

POSITIVE LOGITS

 drawing

0.87

 Drawing

0.86

cards

0.84

Draw

0.83

draw

0.83

 drawn

0.81

 Sketch

0.81

Artist

0.81

 Photoshop

0.80

 drawings

0.79

Activations Density 0.018%

high activation values for the word "draw."

instances of the word "draw" and its variations

No Comments

No Known Activations