INDEX

Explanations

numbers in a specific format

oai_token-act-pair · gpt-3.5-turbo

the term "max" and variations related to maximum values or limits

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

FORE

-0.76

hook

-0.70

 Nare

-0.68

 Forth

-0.67

 Rite

-0.66

velt

-0.65

GROUND

-0.64

 Kurd

-0.64

 Journalism

-0.64

 Kong

-0.63

POSITIVE LOGITS

imus

1.21

imize

1.10

nesota

1.08

imates

1.03

ima

1.01

icum

0.95

ims

0.86

illary

0.85

egal

0.83

imil

0.83

Activations Density 0.012%

numbers in a specific format

the term "max" and variations related to maximum values or limits

No Comments

No Known Activations