INDEX

Explanations

proper nouns, particularly names like "Boris"

oai_token-act-pair · gpt-3.5-turbo

the name "Boris" and its variations in the context of political discussions

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

prise

-0.71

venge

-0.71

etermination

-0.71

intent

-0.68

mble

-0.66

icago

-0.65

imony

-0.65

ppelin

-0.65

WORK

-0.65

ocrine

-0.65

POSITIVE LOGITS

 Boris

1.12

Yel

1.00

ovich

0.93

Dia

0.90

 Johnson

0.86

Nem

0.86

achev

0.83

sov

0.83

 Bere

0.82

 Karl

0.81

Activations Density 0.012%

proper nouns, particularly names like "Boris"

the name "Boris" and its variations in the context of political discussions

No Comments

No Known Activations