INDEX

Explanations

mentions of the word "Victorian"

oai_token-act-pair · gpt-3.5-turbo

references to the Victorian era and associated terms

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

eks

-0.75

anks

-0.73

anka

-0.72

aepernick

-0.72

Sty

-0.71

Bol

-0.67

ucle

-0.67

Teg

-0.67

rounded

-0.66

tem

-0.66

POSITIVE LOGITS

 Victorian

2.59

Alice

1.23

 Alice

1.22

 Fairy

1.16

 Derby

1.04

UP

1.03

 Wonderland

1.02

 Dickens

0.94

 Salem

0.94

Victoria

0.93

Activations Density 0.025%

mentions of the word "Victorian"

references to the Victorian era and associated terms

No Comments

No Known Activations