INDEX

Explanations

the word "nobody" followed by statements suggesting individual actions or characteristics

oai_token-act-pair · gpt-3.5-turbo

references to the word "nobody."

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 expressed

-0.65

 activ

-0.63

 enrichment

-0.62

 indicated

-0.61

 mounts

-0.60

 select

-0.60

 plates

-0.60

 arts

-0.60

 forms

-0.59

app

-0.59

POSITIVE LOGITS

 Nobody

3.26

Nobody

2.31

 Everybody

1.89

 Somebody

1.69

 nobody

1.68

Everybody

1.68

 Everyone

1.53

 Someone

1.33

 Whoever

1.32

 Nothing

1.32

Activations Density 0.012%

the word "nobody" followed by statements suggesting individual actions or characteristics

references to the word "nobody."

No Comments

No Known Activations