INDEX

Explanations

the name "Aki" at varying levels of activation

oai_token-act-pair · gpt-3.5-turbo

mentions of a particular individual, likely a character or public figure, throughout the document

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

child

-0.93

liness

-0.85

craft

-0.79

 unfocusedRange

-0.75

cru

-0.70

advertisement

-0.70

actionDate

-0.69

 Cosponsors

-0.68

bucks

-0.67

egg

-0.67

POSITIVE LOGITS

wi

0.91

ya

0.87

utenant

0.84

ota

0.83

aki

0.82

Ã±a

0.81

uten

0.81

uchi

0.79

uyomi

0.77

pta

0.77

Activations Density 0.019%

the name "Aki" at varying levels of activation

mentions of a particular individual, likely a character or public figure, throughout the document

No Comments

No Known Activations