INDEX

Explanations

prompts or calls to action, often starting with "Want"

oai_token-act-pair · gpt-3.5-turbo

the repeated phrase "Want" indicating desire or request for action

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

è¦ļéĨĴ

-0.83

cise

-0.75

é¾

-0.71

displayText

-0.67

eding

-0.66

workers

-0.65

cision

-0.64

effect

-0.64

urgy

-0.63

sequ

-0.62

POSITIVE LOGITS

 Want

1.00

Want

0.95

 wanna

0.75

MSN

0.72

want

0.71

ickr

0.69

ickets

0.68

reprene

0.68

Us

0.67

awaru

0.66

Activations Density 0.011%

prompts or calls to action, often starting with "Want"

the repeated phrase "Want" indicating desire or request for action

No Comments

No Known Activations