INDEX

Explanations

the word "Less" followed by a number

oai_token-act-pair · gpt-3.5-turbo

phrases indicating a low quantity or cost

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

ãĤ±

-0.67

uu

-0.62

 transporter

-0.59

 submitting

-0.58

este

-0.58

oco

-0.57

anka

-0.57

 tricked

-0.56

 smashing

-0.55

 deceive

-0.55

POSITIVE LOGITS

 Less

3.45

Less

3.01

Few

1.38

 less

1.36

 More

1.35

 Greater

1.34

 Lessons

1.33

 lesser

1.24

LESS

1.16

 Reduced

1.15

Activations Density 0.007%

the word "Less" followed by a number

phrases indicating a low quantity or cost

No Comments

No Known Activations