INDEX

Explanations

instances where people hold certain opinions or make particular claims

oai_token-act-pair · gpt-3.5-turbo

the word "have."

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

eem

-0.69

 Apart

-0.62

catentry

-0.61

—-

-0.61

TG

-0.60

housing

-0.58

dc

-0.57

icking

-0.57

 Anyway

-0.57

 thief

-0.55

POSITIVE LOGITS

 been

1.45

been

1.23

 undergone

1.09

 begun

1.02

 Been

1.00

 become

0.98

 gotten

0.97

 gone

0.96

 arisen

0.94

 seen

0.93

Activations Density 0.336%

instances where people hold certain opinions or make particular claims

the word "have."

No Comments

No Known Activations