INDEX

Explanations

mentions of the word "sister" at high activations

oai_token-act-pair · gpt-3.5-turbo

references to siblings, specifically sisters

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

veyard

-0.74

ered

-0.68

ustomed

-0.67

Frames

-0.67

ocalypse

-0.63

tarians

-0.63

ankind

-0.62

upuncture

-0.61

urations

-0.61

atility

-0.61

POSITIVE LOGITS

hood

1.11

 sister

0.94

hips

0.92

heses

0.89

 sisters

0.79

hesis

0.79

folk

0.79

fax

0.75

 aunt

0.73

 Sister

0.73

Activations Density 0.013%

mentions of the word "sister" at high activations

references to siblings, specifically sisters

No Comments

No Known Activations