INDEX

Explanations

instances of the word "ara" followed by a number, possibly names or codes

oai_token-act-pair · gpt-3.5-turbo

repeated occurrences of the term "ara" in various contexts

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

COP

-0.70

 gentlemen

-0.69

 overse

-0.66

FDR

-0.64

 Nazis

-0.64

 Communists

-0.64

pow

-0.63

 supremacists

-0.62

 analy

-0.62

GPL

-0.61

POSITIVE LOGITS

ara

4.31

ARA

2.32

aran

1.72

arah

1.70

ari

1.55

aris

1.44

aro

1.42

aria

1.37

aru

1.33

ira

1.33

Activations Density 0.011%

instances of the word "ara" followed by a number, possibly names or codes

repeated occurrences of the term "ara" in various contexts

No Comments

No Known Activations