INDEX

Explanations

phrases related to suspicious or illegal activities

oai_token-act-pair · gpt-3.5-turbo

mentions of law enforcement or military organizations, particularly the Department of Defense (DOD)

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ENDED

-0.85

owered

-0.74

 ILCS

-0.73

 Participation

-0.69

MQ

-0.68

 Invest

-0.67

izu

-0.66

 Presence

-0.66

 Protector

-0.66

RESULTS

-0.66

POSITIVE LOGITS

dod

1.12

gers

0.93

bys

0.92

ny

0.89

icles

0.88

gement

0.84

ricks

0.84

esty

0.84

yssey

0.79

yright

0.79

Activations Density 0.017%

phrases related to suspicious or illegal activities

mentions of law enforcement or military organizations, particularly the Department of Defense (DOD)

No Comments

No Known Activations