INDEX

Explanations

Proper nouns related to individuals named "Steve."

oai_token-act-pair · gpt-3.5-turbo

references to the name "Steve."

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

 resolves

-0.78

fet

-0.78

ihar

-0.71

âĶĢ

-0.70

rans

-0.69

mun

-0.69

nex

-0.69

 transit

-0.69

binding

-0.68

 resolved

-0.66

POSITIVE LOGITS

 Steve

3.24

Steve

2.94

 Steven

1.86

 Stephen

1.68

 Tony

1.54

 Mike

1.52

 Pete

1.50

 Chris

1.50

 Kevin

1.49

Steven

1.47

Activations Density 0.019%

Proper nouns related to individuals named "Steve."

references to the name "Steve."

No Comments

No Known Activations