INDEX

Explanations

the word "former" followed by a noun, potentially indicating a comparison or transition

oai_token-act-pair · gpt-3.5-turbo

references to individuals who previously held specific positions or titles

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

oked

-0.82

ographed

-0.79

andise

-0.77

ourced

-0.77

lied

-0.74

opped

-0.71

anche

-0.70

antics

-0.70

owitz

-0.70

anguage

-0.70

POSITIVE LOGITS

 Yugoslavia

1.23

 Yugoslav

1.08

 Soviet

0.92

KGB

0.83

 USSR

0.83

 smoker

0.81

 employer

0.78

 president

0.75

 communist

0.74

 convict

0.73

Activations Density 0.032%

the word "former" followed by a noun, potentially indicating a comparison or transition

references to individuals who previously held specific positions or titles

No Comments

No Known Activations