INDEX

Explanations

the action of being contacted or contacting someone

oai_token-act-pair · gpt-3.5-turbo

instances of the word "contacted"

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

owl

-0.63

ows

-0.62

 deficit

-0.60

 eternity

-0.60

 exposition

-0.60

 style

-0.59

 cheap

-0.58

 Latin

-0.58

 superiority

-0.57

 performance

-0.57

POSITIVE LOGITS

 contacted

3.50

 alerted

1.90

 approached

1.88

 contacting

1.88

 notified

1.88

 consulted

1.74

 emailed

1.72

 interviewed

1.70

 contact

1.49

Contact

1.47

Activations Density 0.014%

the action of being contacted or contacting someone

instances of the word "contacted"

No Comments

No Known Activations