INDEX

Explanations

strongly worded expressions or emphasis, such as 'damn' or 'darn'

oai_token-act-pair · gpt-3.5-turbo

expressions of strong frustration or emphasis

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

NetMessage

-1.10

KY

-0.76

ramid

-0.75

Interstitial

-0.75

KER

-0.72

CRE

-0.71

idon

-0.70

 membr

-0.68

cn

-0.68

chn

-0.68

POSITIVE LOGITS

 darn

0.86

 damn

0.83

selves

0.83

ibly

0.79

 damned

0.76

holes

0.73

ation

0.72

 kidding

0.71

wit

0.71

nuts

0.70

Activations Density 0.018%

strongly worded expressions or emphasis, such as 'damn' or 'darn'

expressions of strong frustration or emphasis

No Comments

No Known Activations