INDEX

Explanations

phrases related to packaging, bundling, and boxing items

oai_token-act-pair · gpt-3.5-turbo

terms related to packaged or bundled goods

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

bia

-0.79

rence

-0.75

reddit

-0.75

rences

-0.74

verson

-0.74

reditary

-0.73

antine

-0.72

nesota

-0.72

pen

-0.71

alter

-0.70

POSITIVE LOGITS

 packaged

1.49

 bundled

1.30

 boxed

0.87

 bund

0.83

 barg

0.80

 bundle

0.79

 bottled

0.78

 bundles

0.75

 packaging

0.74

 separately

0.74

Activations Density 0.007%

phrases related to packaging, bundling, and boxing items

terms related to packaged or bundled goods

No Comments

No Known Activations