INDEX

Explanations

names of individuals mentioned in text

oai_token-act-pair · gpt-3.5-turbo

instances of the conjunction "and"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

root

-0.68

Fuck

-0.63

choice

-0.63

oppable

-0.62

ducers

-0.62

ilt

-0.61

udding

-0.61

ribe

-0.60

ail

-0.60

>[

-0.59

POSITIVE LOGITS

 underwent

1.51

 stayed

1.37

 remained

1.36

 became

1.36

was

1.35

 traveled

1.32

 participated

1.32

 flew

1.31

 wandered

1.30

 withdrew

1.26

Activations Density 0.262%

names of individuals mentioned in text

instances of the conjunction "and"

No Comments

No Known Activations