INDEX

Explanations

words related to restrictions or censorship

oai_token-act-pair · gpt-3.5-turbo

instances of the word "blocked."

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

rouse

-0.78

arist

-0.72

alt

-0.68

rial

-0.68

negie

-0.68

 Institution

-0.68

oi

-0.67

Truth

-0.67

MAT

-0.67

enger

-0.67

POSITIVE LOGITS

 blocked

0.91

 blocking

0.89

abl

0.73

 cancell

0.71

 aback

0.70

 obstruction

0.69

luaj

0.69

cffffcc

0.69

picking

0.68

 blocks

0.67

Activations Density 0.011%

words related to restrictions or censorship

instances of the word "blocked."

No Comments

No Known Activations