INDEX

Explanations

words related to orders or commands

oai_token-act-pair · gpt-3.5-turbo

instances of the word "cord" and its variations

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ãĥ¼ãĤ¯

-0.71

 expired

-0.69

 refresh

-0.69

 slideshow

-0.69

 slides

-0.69

 shaved

-0.67

 chips

-0.66

 popcorn

-0.65

 Akira

-0.64

 steal

-0.63

POSITIVE LOGITS

ord

4.67

ords

2.74

ORD

2.24

ordinate

2.00

orde

1.57

ording

1.55

ordes

1.37

orm

1.33

Ord

1.30

coord

1.26

Activations Density 0.010%

words related to orders or commands

instances of the word "cord" and its variations

No Comments

No Known Activations