INDEX

Explanations

phrases describing capabilities or personal attributes of individuals

oai_token-act-pair · gpt-3.5-turbo

phrases that express admiration or respect for individuals

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

pages

-0.74

cats

-0.74

suits

-0.72

Frames

-0.72

estyles

-0.70

icals

-0.70

pees

-0.69

 Rooms

-0.68

ITS

-0.68

acements

-0.68

POSITIVE LOGITS

 thinker

1.39

 educator

1.37

 lover

1.35

 performer

1.30

 negotiator

1.28

 lecturer

1.27

 protector

1.27

 supporter

1.24

 healer

1.22

 mentor

1.20

Activations Density 0.203%

phrases describing capabilities or personal attributes of individuals

phrases that express admiration or respect for individuals

No Comments

No Known Activations