INDEX

Explanations

the word "whichever"

oai_token-act-pair · gpt-3.5-turbo

phrases indicating choice or preference

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

 Nationwide

-0.75

spect

-0.73

gers

-0.71

gling

-0.69

se

-0.69

ken

-0.68

es

-0.67

 Tome

-0.67

en

-0.67

Se

-0.66

POSITIVE LOGITS

ichever

1.31

soever

1.00

 whichever

0.94

 comr

0.82

wcsstore

0.81

é¾įå¥ĳå£«

0.81

ĵĺ

0.80

 scraps

0.79

DragonMagazine

0.76

oshop

0.76

Activations Density 0.006%

the word "whichever"

phrases indicating choice or preference

No Comments

No Known Activations