INDEX

Explanations

words related to control and authority

oai_token-act-pair · gpt-3.5-turbo

instances of the word "control" and its context, indicating concepts related to power and authority

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

uum

-0.74

tom

-0.68

Mich

-0.67

osp

-0.67

hiba

-0.67

ople

-0.65

ithing

-0.64

istle

-0.64

hire

-0.64

yond

-0.63

POSITIVE LOGITS

 levers

0.84

 controlled

0.82

ively

0.79

 steering

0.78

eering

0.76

 contro

0.75

 controlling

0.71

 control

0.68

 Controlled

0.68

 captive

0.66

Activations Density 0.037%

words related to control and authority

instances of the word "control" and its context, indicating concepts related to power and authority

No Comments

No Known Activations