INDEX

Explanations

the word "all."

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

•All the Features!

No Comments

Negative Logits

ĸļ

-0.61

²¾

-0.60

alid

-0.59

 Delicious

-0.57

 Tall

-0.57

NOT

-0.56

fed

-0.56

 NEXT

-0.55

 Daughter

-0.55

Ri

-0.55

POSITIVE LOGITS

ones

0.82

hetti

0.79

urers

0.72

teness

0.69

oor

0.68

 skeptics

0.66

ardy

0.66

oba

0.66

asio

0.65

urgy

0.65

Activations Density 0.066%

the word "all."

No Comments

No Known Activations