INDEX

Explanations

sequences that resemble numerical or code patterns

oai_token-act-pair · gpt-3.5-turbo

numerical data or statistics

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 nomine

-0.83

annis

-0.75

 interf

-0.73

ccording

-0.73

 neighb

-0.72

 corrid

-0.71

 contrace

-0.71

 referen

-0.71

 horizont

-0.68

 slightest

-0.68

POSITIVE LOGITS

 partName

1.03

 attRot

0.84

)].

0.76

0.70

çļĦ

0.69

0.66

bu

0.64

df

0.64

åĽ

0.63

Activations Density 0.068%

sequences that resemble numerical or code patterns

numerical data or statistics

No Comments

No Known Activations