INDEX

Explanations

phrases related to expressing concerns or worries

oai_token-act-pair · gpt-3.5-turbo

mentions of concerns and issues related to health and safety

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

 Doodle

-0.81

 precincts

-0.78

 dances

-0.76

aunder

-0.73

heid

-0.73

azon

-0.72

OUP

-0.70

 equivalents

-0.70

å§

-0.70

haw

-0.66

POSITIVE LOGITS

 undue

0.80

cens

0.78

severe

0.77

 excessive

0.74

 Patients

0.70

 prolonged

0.69

 premature

0.68

 preventing

0.67

 misinformation

0.67

 heightened

0.66

Activations Density 0.151%

phrases related to expressing concerns or worries

mentions of concerns and issues related to health and safety

No Comments

No Known Activations