INDEX

Explanations

terms related to admiration or reverence towards someone or something

oai_token-act-pair · gpt-3.5-turbo

references to "idols" or figures of admiration and worship

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

aunder

-0.77

 Dull

-0.77

RAW

-0.74

-0.72

--------------------------------------------------------

-0.70

~~~~~~~~~~~~~~~~

-0.68

Klu

-0.67

utenberg

-0.66

xp

-0.66

 Attention

-0.65

POSITIVE LOGITS

 idol

1.32

 idols

1.05

 Idol

0.94

 sacrific

0.94

 worshipped

0.93

ãħĭãħĭ

0.89

ãħĭ

0.83

 worship

0.82

 worsh

0.78

zeb

0.76

Activations Density 0.007%

terms related to admiration or reverence towards someone or something

references to "idols" or figures of admiration and worship

No Comments

No Known Activations