INDEX

Explanations

numbers and special characters that might be associated with sensitive information or data

oai_token-act-pair · gpt-3.5-turbo

special characters or symbols

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 Charlottesville

-0.73

 mole

-0.69

 dividing

-0.67

cko

-0.66

 Gutenberg

-0.65

 clipboard

-0.64

 Mueller

-0.64

 licensee

-0.64

 racists

-0.62

 suprem

-0.62

POSITIVE LOGITS

Ĥ

1.37

¹

1.03

¨

0.97

£

0.97

çĶ°

0.95

Ø§ÙĦ

0.92

atar

0.91

ading

0.91

agara

0.90

Ø¨

0.90

Activations Density 0.006%

numbers and special characters that might be associated with sensitive information or data

special characters or symbols

No Comments

No Known Activations