INDEX

Explanations

phrases related to providing information or instructions to the user

oai_token-act-pair · gpt-3.5-turbo

references to the reader or listener

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

ftime

-0.79

 Sabha

-0.75

 Fried

-0.66

Tang

-0.66

ĸļ

-0.64

 Ange

-0.64

 Advent

-0.62

 Course

-0.62

 Agriculture

-0.62

adelphia

-0.61

POSITIVE LOGITS

 guys

1.03

're

1.02

tub

0.96

 know

0.89

RS

0.87

 naughty

0.78

've

0.76

 glimpse

0.74

 decide

0.72

 understand

0.71

Activations Density 0.079%

phrases related to providing information or instructions to the user

references to the reader or listener

No Comments

No Known Activations