INDEX

Explanations

phrases related to illicit activities or groups

oai_token-act-pair · gpt-3.5-turbo

terms associated with illicit activities

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

 Thunder

-0.69

 Buffalo

-0.69

 Kitchen

-0.65

 quarters

-0.64

 lumber

-0.62

 dash

-0.62

 chicken

-0.61

 Bravo

-0.61

 toast

-0.60

 Sapphire

-0.58

POSITIVE LOGITS

icit

4.79

icity

1.54

icative

1.24

ications

1.23

icient

1.22

ict

1.18

iciency

1.09

icent

1.07

icter

1.06

icia

1.05

Activations Density 0.011%

phrases related to illicit activities or groups

terms associated with illicit activities

No Known Activations