INDEX

Explanations

keywords starting with "Sw" followed by a number

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Sw."

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

Not in Any Lists

No Comments

Negative Logits

DV

-0.75

 Caesar

-0.72

 Burnett

-0.70

 Destiny

-0.69

fig

-0.68

cul

-0.65

 fallacy

-0.65

 character

-0.65

arc

-0.64

 morale

-0.64

POSITIVE LOGITS

Sw

3.81

sw

2.07

Sw

1.96

SW

1.74

 Swed

1.61

 Sweden

1.51

Ice

1.47

Switch

1.39

Swe

1.38

 Swedish

1.34

Activations Density 0.011%

keywords starting with "Sw" followed by a number

instances of the word "Sw."

No Comments

No Known Activations