INDEX

Explanations

phrases related to authority or control

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "take" 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

 properties

-0.69

 relative

-0.63

 peers

-0.62

GP

-0.61

 youth

-0.61

 bats

-0.61

 drivers

-0.60

 cent

-0.60

 const

-0.60

MpServer

-0.60

POSITIVE LOGITS

take

3.83

taking

1.86

Take

1.25

seek

1.22

make

1.12

shall

1.11

skip

1.10

write

1.08

continue

1.07

send

1.06

Activations Density 0.013%

phrases related to authority or control

occurrences of the word "take" in various contexts

No Comments

No Known Activations