INDEX

Explanations

phrases related to news articles or statements

oai_token-act-pair · gpt-3.5-turbo

instances of structured statements or reports

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

¥µ

-0.68

ĪĴ

-0.62

kees

-0.61

aturdays

-0.61

 endeavor

-0.61

utical

-0.60

 Cipher

-0.59

ĸļ

-0.59

ĺħ

-0.58

igslist

-0.58

POSITIVE LOGITS

Britain

0.96

Asked

0.91

However

0.89

Labour

0.87

Shape

0.85

Speaking

0.85

Scotland

0.81

BBC

0.80

SPONSORED

0.79

Writing

0.79

Activations Density 0.577%

phrases related to news articles or statements

instances of structured statements or reports

No Comments

No Known Activations