INDEX

Explanations

phrases containing the word "da"

oai_token-act-pair · gpt-3.5-turbo

references to the concept of "data."

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

 Wyoming

-0.66

 Pratt

-0.65

relative

-0.65

 Lawrence

-0.64

Poly

-0.64

 Cond

-0.63

PG

-0.60

poly

-0.59

 Patriot

-0.59

aways

-0.58

POSITIVE LOGITS

da

3.99

Da

1.94

Da

1.78

da

1.34

 daemon

1.33

di

1.24

dos

1.24

Ã£o

1.15

fa

1.10

Ã§a

1.07

Activations Density 0.002%

phrases containing the word "da"

references to the concept of "data."

No Comments

No Known Activations