INDEX

Explanations

phrases related to giving advice or warnings

oai_token-act-pair · gpt-3.5-turbo

the contraction "don't."

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

 Penguin

-0.73

 Reloaded

-0.73

 SetTextColor

-0.73

ĪĴ

-0.71

 Colour

-0.71

 Radiation

-0.69

£ı

-0.69

 Defin

-0.68

 Dangerous

-0.66

 Positive

-0.66

POSITIVE LOGITS

 necessarily

0.92

cha

0.89

ween

0.88

ting

0.82

ional

0.80

 anymore

0.80

aper

0.80

ardless

0.79

 know

0.79

rave

0.78

Activations Density 0.081%

phrases related to giving advice or warnings

the contraction "don't."

No Comments

No Known Activations