INDEX

Explanations

proper nouns or names of places and entities

oai_token-act-pair · gpt-3.5-turbo

instances of the word "NEW" followed by various numerical values and contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

ppe

-0.87

stood

-0.76

mop

-0.69

-0.67

minecraft

-0.64

arching

-0.64

osate

-0.64

ãĤ¼

-0.63

agn

-0.63

Reloaded

-0.61

POSITIVE LOGITS

 YORK

1.33

foundland

1.19

bie

1.03

PORT

0.96

 York

0.93

 Orleans

0.91

ARK

0.89

bies

0.89

 Zealand

0.87

CAST

0.84

Activations Density 0.010%

proper nouns or names of places and entities

instances of the word "NEW" followed by various numerical values and contexts

No Comments

No Known Activations