INDEX

Explanations

mentions of the word "Sm" followed by either a numerical value of 9 or 10

oai_token-act-pair · gpt-3.5-turbo

mentions of the "Smoker" brand or related terminology

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ARCH

-0.72

orial

-0.70

aga

-0.68

bishop

-0.66

orus

-0.65

 confir

-0.64

yrinth

-0.63

oire

-0.63

 Cerberus

-0.63

agos

-0.63

POSITIVE LOGITS

Sm

3.68

Sm

2.54

sm

1.89

sm

1.60

 Smoke

1.52

SM

1.44

 Smash

1.38

 Smile

1.37

 Smoking

1.35

Sn

1.24

Activations Density 0.020%

mentions of the word "Sm" followed by either a numerical value of 9 or 10

mentions of the "Smoker" brand or related terminology

No Comments

No Known Activations