INDEX

Explanations

words related to security breaches or violations

oai_token-act-pair · gpt-3.5-turbo

instances of the prefix "un" indicating negation or the absence of something

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

 fret

-0.77

 twist

-0.74

 shutter

-0.74

 Lose

-0.73

 madness

-0.71

 Duchess

-0.71

 parting

-0.71

 Veil

-0.70

 lashes

-0.70

 strugg

-0.70

POSITIVE LOGITS

licensed

1.47

authorized

1.44

approved

1.40

author

1.34

verified

1.32

registered

1.31

anticipated

1.23

ruly

1.21

planned

1.21

productive

1.20

Activations Density 0.066%

words related to security breaches or violations

instances of the prefix "un" indicating negation or the absence of something

No Comments

No Known Activations