INDEX

Explanations

the name "Kap" with varying suffixes

oai_token-act-pair · gpt-3.5-turbo

mentions of specific individuals, particularly the name "Kapuscinski"

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

 console

-0.70

fig

-0.64

='

-0.63

Console

-0.62

fell

-0.60

fig

-0.60

 control

-0.57

elf

-0.56

bread

-0.56

 rubber

-0.55

POSITIVE LOGITS

Kap

4.33

Kw

1.37

KP

1.32

 Kaepernick

1.31

 Kund

1.27

Kam

1.24

 Kamp

1.20

 Shak

1.19

 Rath

1.17

Kah

1.14

Activations Density 0.025%

the name "Kap" with varying suffixes

mentions of specific individuals, particularly the name "Kapuscinski"

No Comments

No Known Activations