INDEX

Explanations

phrases related to instructions or warnings

oai_token-act-pair · gpt-3.5-turbo

contractions and negative forms of the verb "to do."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 Pegasus

-0.74

 Craigslist

-0.72

 booked

-0.69

 ejected

-0.65

 sibling

-0.65

 Hercules

-0.63

 matured

-0.63

 Rhodes

-0.62

 wheels

-0.62

 booted

-0.61

POSITIVE LOGITS

âĢ

1.96

 âĢ

1.46

ãĢ

1.37

Ì¶

1.36

Â

1.35

Â§

1.34

îĢ

1.34

âĢł

1.33

ÃĥÃĤ

1.30

âĶ

1.29

Activations Density 0.689%

phrases related to instructions or warnings

contractions and negative forms of the verb "to do."

No Comments

No Known Activations