INDEX

Explanations

phrases related to a specific name or title, potentially "Latta"

oai_token-act-pair · gpt-3.5-turbo

words and phrases associated with identity or recognition, particularly names and titles

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

 shelves

-0.80

 disob

-0.79

 tread

-0.73

 Schwar

-0.69

bows

-0.68

 Valiant

-0.68

 bestselling

-0.67

 Cind

-0.66

 freel

-0.65

jew

-0.64

POSITIVE LOGITS

atta

3.38

rought

1.41

aji

1.31

 Description

1.10

atti

1.05

abba

0.98

TION

0.95

ulla

0.89

houn

0.84

 Amen

0.83

Activations Density 0.020%

phrases related to a specific name or title, potentially "Latta"

words and phrases associated with identity or recognition, particularly names and titles

No Comments

No Known Activations