INDEX

Explanations

words related to descriptions or objects that one can own or possess

oai_token-act-pair · gpt-3.5-turbo

references to locations or origins associated with a variety of subjects

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

ratulations

-0.81

few

-0.71

idav

-0.70

seek

-0.70

merce

-0.70

faced

-0.70

hai

-0.70

lied

-0.69

icators

-0.68

threat

-0.67

POSITIVE LOGITS

 afar

1.37

 whence

1.09

 thence

0.86

 anywhere

0.80

 abroad

0.80

Uni

0.79

 somewhere

0.78

 Epic

0.75

 scratch

0.74

 underneath

0.74

Activations Density 0.140%

words related to descriptions or objects that one can own or possess

references to locations or origins associated with a variety of subjects

No Comments

No Known Activations