INDEX

Explanations

phrases indicating surprise or contradiction

oai_token-act-pair · gpt-3.5-turbo

the word "actually" in various contexts

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

lain

-0.73

illed

-0.69

fu

-0.69

bye

-0.68

wich

-0.68

cit

-0.67

oute

-0.66

heid

-0.63

legged

-0.63

ado

-0.63

POSITIVE LOGITS

 comprom

0.88

 meant

0.83

 bothering

0.81

olkien

0.80

 REALLY

0.77

 metic

0.73

 bother

0.72

 quite

0.71

 intended

0.68

 actually

0.67

Activations Density 0.024%

phrases indicating surprise or contradiction

the word "actually" in various contexts

No Comments

No Known Activations