INDEX

Explanations

mentions of individuals being given positions of authority or recognition by others

oai_token-act-pair · gpt-3.5-turbo

instances of the phrase "by" indicating authorship or attribution

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Negative Logits

ational

-0.72

BSD

-0.68

 "$:/

-0.68

itto

-0.66

bian

-0.64

amera

-0.64

asy

-0.63

Finish

-0.63

ginx

-0.63

idad

-0.63

POSITIVE LOGITS

 virtue

1.35

laws

1.06

products

0.97

gone

0.92

 fiat

0.83

 multiplying

0.80

 default

0.76

 leaps

0.75

 proxy

0.75

akuya

0.71

Activations Density 0.153%

mentions of individuals being given positions of authority or recognition by others

instances of the phrase "by" indicating authorship or attribution

No Known Activations